deit-training-oulu.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "76ff1da1",
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_kg_hide-output": true,
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
    "execution": {
     "iopub.execute_input": "2023-03-25T15:29:31.699852Z",
     "iopub.status.busy": "2023-03-25T15:29:31.698906Z",
     "iopub.status.idle": "2023-03-25T15:30:26.650984Z",
     "shell.execute_reply": "2023-03-25T15:30:26.649604Z"
    },
    "papermill": {
     "duration": 54.965224,
     "end_time": "2023-03-25T15:30:26.654168",
     "exception": false,
     "start_time": "2023-03-25T15:29:31.688944",
     "status": "completed"
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: timm in /opt/conda/lib/python3.7/site-packages (0.6.12)\r\n",
      "Requirement already satisfied: torch>=1.7 in /opt/conda/lib/python3.7/site-packages (from timm) (1.13.0)\r\n",
      "Requirement already satisfied: huggingface-hub in /opt/conda/lib/python3.7/site-packages (from timm) (0.12.1)\r\n",
      "Requirement already satisfied: torchvision in /opt/conda/lib/python3.7/site-packages (from timm) (0.14.0)\r\n",
      "Requirement already satisfied: pyyaml in /opt/conda/lib/python3.7/site-packages (from timm) (6.0)\r\n",
      "Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from torch>=1.7->timm) (4.4.0)\r\n",
      "Requirement already satisfied: packaging>=20.9 in /opt/conda/lib/python3.7/site-packages (from huggingface-hub->timm) (23.0)\r\n",
      "Requirement already satisfied: filelock in /opt/conda/lib/python3.7/site-packages (from huggingface-hub->timm) (3.9.0)\r\n",
      "Requirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from huggingface-hub->timm) (4.11.4)\r\n",
      "Requirement already satisfied: requests in /opt/conda/lib/python3.7/site-packages (from huggingface-hub->timm) (2.28.2)\r\n",
      "Requirement already satisfied: tqdm>=4.42.1 in /opt/conda/lib/python3.7/site-packages (from huggingface-hub->timm) (4.64.1)\r\n",
      "Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /opt/conda/lib/python3.7/site-packages (from torchvision->timm) (9.4.0)\r\n",
      "Requirement already satisfied: numpy in /opt/conda/lib/python3.7/site-packages (from torchvision->timm) (1.21.6)\r\n",
      "Requirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->huggingface-hub->timm) (3.11.0)\r\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests->huggingface-hub->timm) (2022.12.7)\r\n",
      "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests->huggingface-hub->timm) (1.26.14)\r\n",
      "Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests->huggingface-hub->timm) (3.4)\r\n",
      "Requirement already satisfied: charset-normalizer<4,>=2 in /opt/conda/lib/python3.7/site-packages (from requests->huggingface-hub->timm) (2.1.1)\r\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\r\n",
      "\u001b[0mCollecting oulumetrics\r\n",
      "  Downloading oulumetrics-0.1.tar.gz (1.8 kB)\r\n",
      "  Preparing metadata (setup.py) ... \u001b[?25l-\b \bdone\r\n",
      "\u001b[?25hBuilding wheels for collected packages: oulumetrics\r\n",
      "  Building wheel for oulumetrics (setup.py) ... \u001b[?25l-\b \b\\\b \bdone\r\n",
      "\u001b[?25h  Created wheel for oulumetrics: filename=oulumetrics-0.1-py3-none-any.whl size=2459 sha256=e5b158d7baab8da47b83ebb7a395d8863040e6fe3fb0b4d192aac75336bba965\r\n",
      "  Stored in directory: /root/.cache/pip/wheels/7a/12/a3/9f8ea1ed198c858c48af6153e447a4c474f4f79c9b82abe104\r\n",
      "Successfully built oulumetrics\r\n",
      "Installing collected packages: oulumetrics\r\n",
      "Successfully installed oulumetrics-0.1\r\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\r\n",
      "\u001b[0mRequirement already satisfied: keras in /opt/conda/lib/python3.7/site-packages (2.11.0)\r\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\r\n",
      "\u001b[0mRequirement already satisfied: tensorflow-addons in /opt/conda/lib/python3.7/site-packages (0.19.0)\r\n",
      "Requirement already satisfied: packaging in /opt/conda/lib/python3.7/site-packages (from tensorflow-addons) (23.0)\r\n",
      "Requirement already satisfied: typeguard>=2.7 in /opt/conda/lib/python3.7/site-packages (from tensorflow-addons) (2.13.3)\r\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\r\n",
      "\u001b[0mCollecting gdown\r\n",
      "  Downloading gdown-4.7.1-py3-none-any.whl (15 kB)\r\n",
      "Requirement already satisfied: beautifulsoup4 in /opt/conda/lib/python3.7/site-packages (from gdown) (4.11.1)\r\n",
      "Requirement already satisfied: tqdm in /opt/conda/lib/python3.7/site-packages (from gdown) (4.64.1)\r\n",
      "Requirement already satisfied: six in /opt/conda/lib/python3.7/site-packages (from gdown) (1.16.0)\r\n",
      "Requirement already satisfied: filelock in /opt/conda/lib/python3.7/site-packages (from gdown) (3.9.0)\r\n",
      "Requirement already satisfied: requests[socks] in /opt/conda/lib/python3.7/site-packages (from gdown) (2.28.2)\r\n",
      "Requirement already satisfied: soupsieve>1.2 in /opt/conda/lib/python3.7/site-packages (from beautifulsoup4->gdown) (2.3.2.post1)\r\n",
      "Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests[socks]->gdown) (3.4)\r\n",
      "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests[socks]->gdown) (1.26.14)\r\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests[socks]->gdown) (2022.12.7)\r\n",
      "Requirement already satisfied: charset-normalizer<4,>=2 in /opt/conda/lib/python3.7/site-packages (from requests[socks]->gdown) (2.1.1)\r\n",
      "Requirement already satisfied: PySocks!=1.5.7,>=1.5.6 in /opt/conda/lib/python3.7/site-packages (from requests[socks]->gdown) (1.7.1)\r\n",
      "Installing collected packages: gdown\r\n",
      "Successfully installed gdown-4.7.1\r\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\r\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import os\n",
    "import torch\n",
    "import torchvision\n",
    "from torchvision import datasets\n",
    "from torchvision import transforms as T # for simplifying the transforms\n",
    "from torch import nn, optim\n",
    "from torch.nn import functional as F\n",
    "from torch.utils.data import DataLoader, sampler, random_split\n",
    "from torchvision import models\n",
    "## Now, we import timm, torchvision image models\n",
    "!pip install timm # kaggle doesnt have it installed by default\n",
    "import timm\n",
    "from timm.loss import LabelSmoothingCrossEntropy # This is better than normal nn.CrossEntropyLoss\n",
    "# remove warnings\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import sys\n",
    "from tqdm import tqdm\n",
    "import time\n",
    "import copy\n",
    "import sklearn.metrics\n",
    "from sklearn.metrics import confusion_matrix, f1_score\n",
    "!pip install oulumetrics\n",
    "import oulumetrics\n",
    "!pip install keras\n",
    "!pip install tensorflow-addons\n",
    "!pip install gdown\n",
    "import subprocess"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "02bb3306",
   "metadata": {
    "papermill": {
     "duration": 0.008357,
     "end_time": "2023-03-25T15:30:26.671764",
     "exception": false,
     "start_time": "2023-03-25T15:30:26.663407",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Dataset Extraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "2c0714c4",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:26.690722Z",
     "iopub.status.busy": "2023-03-25T15:30:26.690383Z",
     "iopub.status.idle": "2023-03-25T15:30:26.695311Z",
     "shell.execute_reply": "2023-03-25T15:30:26.694266Z"
    },
    "papermill": {
     "duration": 0.017206,
     "end_time": "2023-03-25T15:30:26.697387",
     "exception": false,
     "start_time": "2023-03-25T15:30:26.680181",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "protocol = 'Protocol1'\n",
    "color_space = 'RGB'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "c960ba93",
   "metadata": {
    "_kg_hide-output": true,
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:26.715934Z",
     "iopub.status.busy": "2023-03-25T15:30:26.715103Z",
     "iopub.status.idle": "2023-03-25T15:30:36.518469Z",
     "shell.execute_reply": "2023-03-25T15:30:36.517111Z"
    },
    "papermill": {
     "duration": 9.815606,
     "end_time": "2023-03-25T15:30:36.521360",
     "exception": false,
     "start_time": "2023-03-25T15:30:26.705754",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/gdown/cli.py:130: FutureWarning: Option `--id` was deprecated in version 4.3.1 and will be removed in 5.0. You don't need to pass it anymore to use a file ID.\r\n",
      "  category=FutureWarning,\r\n",
      "Downloading...\r\n",
      "From (uriginal): https://drive.google.com/uc?id=1EzhHs0C6UR3-CjN9SAtTUyGI3zRI0tAG\r\n",
      "From (redirected): https://drive.google.com/uc?id=1EzhHs0C6UR3-CjN9SAtTUyGI3zRI0tAG&confirm=t&uuid=f7ecfe66-b7da-42f7-8689-9682a5bfc4a8\r\n",
      "To: /kaggle/working/Protocol1_RGB_Seperated_undersampled.zip\r\n",
      "100%|█████████████████████████████████████████| 451M/451M [00:02<00:00, 172MB/s]\r\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "CompletedProcess(args=['unzip', '-q', 'Protocol1_RGB_Seperated_undersampled'], returncode=0)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "!gdown --id 1EzhHs0C6UR3-CjN9SAtTUyGI3zRI0tAG\n",
    "\n",
    "\n",
    "# specify the file name and destination directory\n",
    "file_name = f'{protocol}_{color_space}_Seperated_undersampled'\n",
    "# construct the command to unzip the file\n",
    "command = ['unzip','-q', file_name]\n",
    "# execute the command\n",
    "subprocess.run(command)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25062b4c",
   "metadata": {
    "papermill": {
     "duration": 0.015592,
     "end_time": "2023-03-25T15:30:36.553017",
     "exception": false,
     "start_time": "2023-03-25T15:30:36.537425",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Dataset Loading"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f7ca788e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:36.584741Z",
     "iopub.status.busy": "2023-03-25T15:30:36.583936Z",
     "iopub.status.idle": "2023-03-25T15:30:36.589887Z",
     "shell.execute_reply": "2023-03-25T15:30:36.588857Z"
    },
    "papermill": {
     "duration": 0.02636,
     "end_time": "2023-03-25T15:30:36.594209",
     "exception": false,
     "start_time": "2023-03-25T15:30:36.567849",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "dataset_path = f\"/kaggle/working/{file_name}\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2b36f22c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:36.629262Z",
     "iopub.status.busy": "2023-03-25T15:30:36.628934Z",
     "iopub.status.idle": "2023-03-25T15:30:36.644233Z",
     "shell.execute_reply": "2023-03-25T15:30:36.643353Z"
    },
    "papermill": {
     "duration": 0.035834,
     "end_time": "2023-03-25T15:30:36.646969",
     "exception": false,
     "start_time": "2023-03-25T15:30:36.611135",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def get_data_loaders(data_dir, batch_size, train = False):\n",
    "    if train:\n",
    "        #train\n",
    "        transform = T.Compose([\n",
    "            T.RandomHorizontalFlip(),\n",
    "            T.RandomVerticalFlip(),\n",
    "#             T.RandomApply(torch.nn.ModuleList([T.ColorJitter()]), p=0.25),\n",
    "            T.Resize(256),\n",
    "            T.CenterCrop(224),\n",
    "            T.ToTensor(),\n",
    "            T.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)), # imagenet means\n",
    "            T.RandomErasing(p=0.20, value='random')\n",
    "        ])\n",
    "        train_data = datasets.ImageFolder(os.path.join(data_dir, \"train/\"), transform = transform)\n",
    "        train_loader = DataLoader(train_data, batch_size=batch_size, shuffle=True, num_workers=4)\n",
    "        return train_loader, len(train_data)\n",
    "    else:\n",
    "        # val/test\n",
    "        transform = T.Compose([ # We dont need augmentation for test transforms\n",
    "            T.Resize(256),\n",
    "            T.CenterCrop(224),\n",
    "            T.ToTensor(),\n",
    "            T.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)), # imagenet means\n",
    "        ])\n",
    "        val_data = datasets.ImageFolder(os.path.join(data_dir, \"dev/\"), transform=transform)\n",
    "        test_data = datasets.ImageFolder(os.path.join(data_dir, \"test/\"), transform=transform)\n",
    "        val_loader = DataLoader(val_data, batch_size=batch_size, shuffle=True, num_workers=4)\n",
    "        test_loader = DataLoader(test_data, batch_size=batch_size, shuffle=True, num_workers=4)\n",
    "        return val_loader, test_loader, len(val_data), len(test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "66c63b8c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:36.699276Z",
     "iopub.status.busy": "2023-03-25T15:30:36.698947Z",
     "iopub.status.idle": "2023-03-25T15:30:36.704804Z",
     "shell.execute_reply": "2023-03-25T15:30:36.703738Z"
    },
    "papermill": {
     "duration": 0.028724,
     "end_time": "2023-03-25T15:30:36.707575",
     "exception": false,
     "start_time": "2023-03-25T15:30:36.678851",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def get_classes(data_dir):\n",
    "    all_data = datasets.ImageFolder(data_dir)\n",
    "    return all_data.classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "931e1dd0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:36.742029Z",
     "iopub.status.busy": "2023-03-25T15:30:36.741741Z",
     "iopub.status.idle": "2023-03-25T15:30:36.880438Z",
     "shell.execute_reply": "2023-03-25T15:30:36.879237Z"
    },
    "papermill": {
     "duration": 0.158111,
     "end_time": "2023-03-25T15:30:36.883267",
     "exception": false,
     "start_time": "2023-03-25T15:30:36.725156",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "(train_loader, train_data_len) = get_data_loaders(dataset_path, 128, train=True)\n",
    "(dev_loader, test_loader, dev_data_len, test_data_len) = get_data_loaders(dataset_path, 32, train=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "ea1e296d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:36.917019Z",
     "iopub.status.busy": "2023-03-25T15:30:36.915797Z",
     "iopub.status.idle": "2023-03-25T15:30:36.966369Z",
     "shell.execute_reply": "2023-03-25T15:30:36.964203Z"
    },
    "papermill": {
     "duration": 0.068824,
     "end_time": "2023-03-25T15:30:36.968667",
     "exception": false,
     "start_time": "2023-03-25T15:30:36.899843",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['fake', 'true'] 2\n"
     ]
    }
   ],
   "source": [
    "classes = get_classes(f\"/kaggle/working/{file_name}/train/\")\n",
    "print(classes, len(classes))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "5487d7b5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:36.990090Z",
     "iopub.status.busy": "2023-03-25T15:30:36.989812Z",
     "iopub.status.idle": "2023-03-25T15:30:36.994782Z",
     "shell.execute_reply": "2023-03-25T15:30:36.993333Z"
    },
    "papermill": {
     "duration": 0.018502,
     "end_time": "2023-03-25T15:30:36.997290",
     "exception": false,
     "start_time": "2023-03-25T15:30:36.978788",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "dataloaders = {\n",
    "    \"train\": train_loader,\n",
    "    \"val\": dev_loader\n",
    "}\n",
    "dataset_sizes = {\n",
    "    \"train\": train_data_len,\n",
    "    \"val\": dev_data_len\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "cfd99573",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:37.018906Z",
     "iopub.status.busy": "2023-03-25T15:30:37.018631Z",
     "iopub.status.idle": "2023-03-25T15:30:37.023578Z",
     "shell.execute_reply": "2023-03-25T15:30:37.022606Z"
    },
    "papermill": {
     "duration": 0.018726,
     "end_time": "2023-03-25T15:30:37.026411",
     "exception": false,
     "start_time": "2023-03-25T15:30:37.007685",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "45 283 243\n"
     ]
    }
   ],
   "source": [
    "print(len(train_loader), len(dev_loader), len(test_loader))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "991cea42",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:37.048633Z",
     "iopub.status.busy": "2023-03-25T15:30:37.047854Z",
     "iopub.status.idle": "2023-03-25T15:30:37.053677Z",
     "shell.execute_reply": "2023-03-25T15:30:37.052361Z"
    },
    "papermill": {
     "duration": 0.018699,
     "end_time": "2023-03-25T15:30:37.055717",
     "exception": false,
     "start_time": "2023-03-25T15:30:37.037018",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5663 9027 7758\n"
     ]
    }
   ],
   "source": [
    "print(train_data_len, dev_data_len, test_data_len)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "57a8e240",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:37.077768Z",
     "iopub.status.busy": "2023-03-25T15:30:37.076980Z",
     "iopub.status.idle": "2023-03-25T15:30:37.142913Z",
     "shell.execute_reply": "2023-03-25T15:30:37.141827Z"
    },
    "papermill": {
     "duration": 0.078951,
     "end_time": "2023-03-25T15:30:37.145109",
     "exception": false,
     "start_time": "2023-03-25T15:30:37.066158",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "device(type='cuda')"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# now, for the model\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "device"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "acd9aae6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:37.167500Z",
     "iopub.status.busy": "2023-03-25T15:30:37.166675Z",
     "iopub.status.idle": "2023-03-25T15:30:37.171200Z",
     "shell.execute_reply": "2023-03-25T15:30:37.170276Z"
    },
    "papermill": {
     "duration": 0.017814,
     "end_time": "2023-03-25T15:30:37.173247",
     "exception": false,
     "start_time": "2023-03-25T15:30:37.155433",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# model_names = timm.list_models(pretrained=True)\n",
    "# print(model_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "559be1b2",
   "metadata": {
    "papermill": {
     "duration": 0.00998,
     "end_time": "2023-03-25T15:30:37.193597",
     "exception": false,
     "start_time": "2023-03-25T15:30:37.183617",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Model Parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "3f5c7c52",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:37.215821Z",
     "iopub.status.busy": "2023-03-25T15:30:37.215013Z",
     "iopub.status.idle": "2023-03-25T15:30:37.219772Z",
     "shell.execute_reply": "2023-03-25T15:30:37.218458Z"
    },
    "papermill": {
     "duration": 0.01833,
     "end_time": "2023-03-25T15:30:37.222112",
     "exception": false,
     "start_time": "2023-03-25T15:30:37.203782",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "pretrained_model_name = 'deit_base_patch16_224'   #cait_M48 cait_XXS36_224 cait_S24_224 deit_base_distilled_patch16_224 deit_base_patch16_224 deit_tiny_distilled_patch16_224\n",
    "no_epochs = 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "461b6edb",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:37.243394Z",
     "iopub.status.busy": "2023-03-25T15:30:37.243096Z",
     "iopub.status.idle": "2023-03-25T15:30:46.798644Z",
     "shell.execute_reply": "2023-03-25T15:30:46.797602Z"
    },
    "papermill": {
     "duration": 9.569035,
     "end_time": "2023-03-25T15:30:46.801180",
     "exception": false,
     "start_time": "2023-03-25T15:30:37.232145",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Downloading: \"https://github.com/facebookresearch/deit/zipball/main\" to /root/.cache/torch/hub/main.zip\n",
      "Downloading: \"https://dl.fbaipublicfiles.com/deit/deit_base_patch16_224-b5f2ef4d.pth\" to /root/.cache/torch/hub/checkpoints/deit_base_patch16_224-b5f2ef4d.pth\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0354f582f33b449a82b70a7aa9766fa1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0.00/330M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "learning_rate = 0.03\n",
    "model = torch.hub.load('facebookresearch/deit:main', pretrained_model_name, pretrained=True)\n",
    "\n",
    "for param in model.parameters(): #freeze model\n",
    "    param.requires_grad = False\n",
    "\n",
    "n_inputs = model.head.in_features\n",
    "model.head = nn.Sequential(\n",
    "    \n",
    "#     nn.BatchNorm1d(192),\n",
    "#     nn.Linear(n_inputs, 512),\n",
    "#     nn.ReLU(),\n",
    "# #     nn.BatchNorm1d(512),\n",
    "#     nn.Dropout(0.2, inplace=True),\n",
    "#     nn.Linear(512, 256),\n",
    "#     nn.ReLU(),\n",
    "#     nn.Dropout(0.2, inplace=True),\n",
    "#     nn.Linear(256, 128),\n",
    "#     nn.ReLU(),\n",
    "#     nn.Dropout(0.2, inplace=True),\n",
    "#     nn.Linear(128, 64),\n",
    "#     nn.ReLU(),\n",
    "# #     nn.Dropout(0.2, inplace=True),\n",
    "    \n",
    "#     nn.Linear(64, len(classes))\n",
    "    ############################################\n",
    "    nn.BatchNorm1d(768),\n",
    "    nn.Linear(n_inputs, 512),\n",
    "    nn.Mish(),\n",
    "#     nn.BatchNorm1d(512),\n",
    "    nn.Dropout(0.2, inplace=True),\n",
    "    nn.Linear(512, 256),\n",
    "    nn.Mish(),\n",
    "    nn.Dropout(0.2, inplace=True),\n",
    "    nn.Linear(256, 128),\n",
    "    nn.Mish(),\n",
    "    nn.Dropout(0.2, inplace=True),\n",
    "    nn.Linear(128, 64),\n",
    "    nn.Mish(),\n",
    "#     nn.Dropout(0.2, inplace=True),\n",
    "    \n",
    "    nn.Linear(64, len(classes))\n",
    "    ##############################################\n",
    "    \n",
    "#     nn.BatchNorm1d(192),\n",
    "# #     nn.Linear(n_inputs, 512),\n",
    "#     nn.PReLU(),\n",
    "# #     nn.BatchNorm1d(512),\n",
    "#     nn.Dropout(0.2, inplace=True),\n",
    "# #     nn.Linear(512, 256),\n",
    "#     nn.PReLU(),\n",
    "#     nn.Dropout(0.2, inplace=True),\n",
    "# #     nn.Linear(256, 128),\n",
    "#     nn.PReLU(),\n",
    "#     nn.Dropout(0.2, inplace=True),\n",
    "# #     nn.Linear(128, 64),\n",
    "#     nn.PReLU(),\n",
    "    #################################\n",
    "#     nn.Linear(n_inputs, 512),\n",
    "#     nn.ReLU(),\n",
    "#     nn.Dropout(0.3),\n",
    "#     nn.Linear(512, len(classes))\n",
    ")\n",
    "\n",
    "# model.head = head_params\n",
    "model = model.to(device)\n",
    "# print(model)\n",
    "# parameters\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "c14db496",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:46.828217Z",
     "iopub.status.busy": "2023-03-25T15:30:46.827835Z",
     "iopub.status.idle": "2023-03-25T15:30:46.834284Z",
     "shell.execute_reply": "2023-03-25T15:30:46.832368Z"
    },
    "papermill": {
     "duration": 0.021828,
     "end_time": "2023-03-25T15:30:46.836609",
     "exception": false,
     "start_time": "2023-03-25T15:30:46.814781",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "criterion = LabelSmoothingCrossEntropy()\n",
    "criterion = criterion.to(device)\n",
    "optimizer = optim.SGD(model.head.parameters(), lr=learning_rate, momentum=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "cee7eae3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:46.860730Z",
     "iopub.status.busy": "2023-03-25T15:30:46.860134Z",
     "iopub.status.idle": "2023-03-25T15:30:46.865235Z",
     "shell.execute_reply": "2023-03-25T15:30:46.864345Z"
    },
    "papermill": {
     "duration": 0.018968,
     "end_time": "2023-03-25T15:30:46.867349",
     "exception": false,
     "start_time": "2023-03-25T15:30:46.848381",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# lr scheduler\n",
    "exp_lr_scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=3, gamma=0.97)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "8f8750e3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:46.890797Z",
     "iopub.status.busy": "2023-03-25T15:30:46.890523Z",
     "iopub.status.idle": "2023-03-25T15:30:46.903575Z",
     "shell.execute_reply": "2023-03-25T15:30:46.902703Z"
    },
    "papermill": {
     "duration": 0.027094,
     "end_time": "2023-03-25T15:30:46.905689",
     "exception": false,
     "start_time": "2023-03-25T15:30:46.878595",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "train_accs = []\n",
    "val_accs = []\n",
    "train_losses = []\n",
    "val_losses=[]\n",
    "\n",
    "def train_model(model, criterion, optimizer, scheduler, num_epochs):\n",
    "    since = time.time()\n",
    "    best_model_wts = copy.deepcopy(model.state_dict())\n",
    "    best_acc = 0.0\n",
    "    \n",
    "    for epoch in range(num_epochs):\n",
    "        print(f'Epoch {epoch}/{num_epochs - 1}')\n",
    "        print(\"-\"*10)\n",
    "        \n",
    "        for phase in ['train', 'val']: # We do training and validation phase per epoch\n",
    "            if phase == 'train':\n",
    "                model.train() # model to training mode\n",
    "            else:\n",
    "                model.eval() # model to evaluate\n",
    "            \n",
    "            running_loss = 0.0\n",
    "            running_corrects = 0.0\n",
    "            \n",
    "            for inputs, labels in tqdm(dataloaders[phase]):\n",
    "                inputs = inputs.to(device)\n",
    "                labels = labels.to(device)\n",
    "                \n",
    "                optimizer.zero_grad()\n",
    "                \n",
    "                with torch.set_grad_enabled(phase == 'train'): # no autograd makes validation go faster\n",
    "                    outputs = model(inputs)\n",
    "#                     print(outputs)\n",
    "                    _, preds = torch.max(outputs,1) # used for accuracy\n",
    "                    loss = criterion(outputs, labels)\n",
    "                    \n",
    "                    if phase == 'train':\n",
    "                        loss.backward()\n",
    "                        optimizer.step()\n",
    "                running_loss += loss.item() * inputs.size(0)\n",
    "                running_corrects += torch.sum(preds == labels.data)\n",
    "                \n",
    "            if phase == 'train':\n",
    "                scheduler.step() # step at end of epoch\n",
    "            \n",
    "            epoch_loss = running_loss / dataset_sizes[phase]\n",
    "            epoch_acc =  running_corrects.double() / dataset_sizes[phase]\n",
    "            epoch_acc = epoch_acc.cpu()\n",
    "            if phase == 'val':\n",
    "                val_accs.append(epoch_acc)\n",
    "                val_losses.append(epoch_loss)\n",
    "            else:\n",
    "                train_accs.append(epoch_acc)\n",
    "                train_losses.append(epoch_loss)\n",
    "            \n",
    "            print(\"{} Loss: {:.4f} Acc: {:.4f}\".format(phase, epoch_loss, epoch_acc))\n",
    "            \n",
    "            if phase == 'val' and epoch_acc > best_acc:\n",
    "                best_acc = epoch_acc\n",
    "                best_model_wts = copy.deepcopy(model.state_dict()) # keep the best validation accuracy model\n",
    "        print()\n",
    "    time_elapsed = time.time() - since # slight error\n",
    "    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))\n",
    "    print(\"Best Train Acc: {:.4f}\".format(max(train_accs)))\n",
    "    print(\"Best Val Acc: {:.4f}\".format(best_acc))\n",
    "    \n",
    "    model.load_state_dict(best_model_wts)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "546839a5",
   "metadata": {
    "papermill": {
     "duration": 0.010678,
     "end_time": "2023-03-25T15:30:46.927619",
     "exception": false,
     "start_time": "2023-03-25T15:30:46.916941",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Model Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "742d988d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T15:30:46.950888Z",
     "iopub.status.busy": "2023-03-25T15:30:46.950282Z",
     "iopub.status.idle": "2023-03-25T16:45:50.501229Z",
     "shell.execute_reply": "2023-03-25T16:45:50.499985Z"
    },
    "papermill": {
     "duration": 4503.566538,
     "end_time": "2023-03-25T16:45:50.504618",
     "exception": false,
     "start_time": "2023-03-25T15:30:46.938080",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 0/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:39<00:00,  1.15it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.6931 Acc: 0.5151\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.07it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.6985 Acc: 0.2651\n",
      "\n",
      "Epoch 1/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.6875 Acc: 0.5547\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.6965 Acc: 0.3703\n",
      "\n",
      "Epoch 2/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.6806 Acc: 0.6405\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.6893 Acc: 0.5513\n",
      "\n",
      "Epoch 3/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.6636 Acc: 0.7081\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.6589 Acc: 0.7326\n",
      "\n",
      "Epoch 4/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.6170 Acc: 0.7597\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.07it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.5649 Acc: 0.8097\n",
      "\n",
      "Epoch 5/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.5170 Acc: 0.8075\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.4718 Acc: 0.8362\n",
      "\n",
      "Epoch 6/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.4456 Acc: 0.8425\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.07it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.4206 Acc: 0.8664\n",
      "\n",
      "Epoch 7/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.4000 Acc: 0.8773\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3909 Acc: 0.8870\n",
      "\n",
      "Epoch 8/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3713 Acc: 0.9011\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3798 Acc: 0.8979\n",
      "\n",
      "Epoch 9/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3491 Acc: 0.9172\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.07it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3848 Acc: 0.8920\n",
      "\n",
      "Epoch 10/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.34it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3368 Acc: 0.9271\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.07it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3736 Acc: 0.9050\n",
      "\n",
      "Epoch 11/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3307 Acc: 0.9315\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3696 Acc: 0.9037\n",
      "\n",
      "Epoch 12/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.34it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3225 Acc: 0.9359\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3618 Acc: 0.9091\n",
      "\n",
      "Epoch 13/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.34it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3104 Acc: 0.9447\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3593 Acc: 0.9131\n",
      "\n",
      "Epoch 14/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3075 Acc: 0.9513\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3649 Acc: 0.9076\n",
      "\n",
      "Epoch 15/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3044 Acc: 0.9509\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3662 Acc: 0.9074\n",
      "\n",
      "Epoch 16/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.3028 Acc: 0.9504\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3461 Acc: 0.9216\n",
      "\n",
      "Epoch 17/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.31it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2969 Acc: 0.9551\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.03it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3337 Acc: 0.9301\n",
      "\n",
      "Epoch 18/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2955 Acc: 0.9574\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3366 Acc: 0.9287\n",
      "\n",
      "Epoch 19/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2932 Acc: 0.9553\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3405 Acc: 0.9233\n",
      "\n",
      "Epoch 20/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2879 Acc: 0.9615\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3449 Acc: 0.9242\n",
      "\n",
      "Epoch 21/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2899 Acc: 0.9597\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3510 Acc: 0.9178\n",
      "\n",
      "Epoch 22/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2839 Acc: 0.9656\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3270 Acc: 0.9318\n",
      "\n",
      "Epoch 23/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.30it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2848 Acc: 0.9606\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3289 Acc: 0.9291\n",
      "\n",
      "Epoch 24/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2805 Acc: 0.9666\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3458 Acc: 0.9213\n",
      "\n",
      "Epoch 25/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2825 Acc: 0.9666\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3296 Acc: 0.9328\n",
      "\n",
      "Epoch 26/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2800 Acc: 0.9675\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3320 Acc: 0.9309\n",
      "\n",
      "Epoch 27/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2745 Acc: 0.9712\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3272 Acc: 0.9334\n",
      "\n",
      "Epoch 28/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2762 Acc: 0.9672\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3257 Acc: 0.9344\n",
      "\n",
      "Epoch 29/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2741 Acc: 0.9695\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3289 Acc: 0.9310\n",
      "\n",
      "Epoch 30/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2764 Acc: 0.9686\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.07it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3404 Acc: 0.9235\n",
      "\n",
      "Epoch 31/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2757 Acc: 0.9689\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3253 Acc: 0.9343\n",
      "\n",
      "Epoch 32/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2734 Acc: 0.9709\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3280 Acc: 0.9320\n",
      "\n",
      "Epoch 33/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2722 Acc: 0.9698\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3256 Acc: 0.9345\n",
      "\n",
      "Epoch 34/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2703 Acc: 0.9726\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3260 Acc: 0.9344\n",
      "\n",
      "Epoch 35/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2673 Acc: 0.9763\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3231 Acc: 0.9354\n",
      "\n",
      "Epoch 36/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2665 Acc: 0.9755\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3230 Acc: 0.9336\n",
      "\n",
      "Epoch 37/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2671 Acc: 0.9733\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3148 Acc: 0.9374\n",
      "\n",
      "Epoch 38/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2667 Acc: 0.9732\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3198 Acc: 0.9366\n",
      "\n",
      "Epoch 39/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2649 Acc: 0.9735\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3127 Acc: 0.9402\n",
      "\n",
      "Epoch 40/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2659 Acc: 0.9751\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3214 Acc: 0.9353\n",
      "\n",
      "Epoch 41/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2625 Acc: 0.9765\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3183 Acc: 0.9383\n",
      "\n",
      "Epoch 42/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.34it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2602 Acc: 0.9781\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3151 Acc: 0.9400\n",
      "\n",
      "Epoch 43/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2597 Acc: 0.9792\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.03it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3167 Acc: 0.9370\n",
      "\n",
      "Epoch 44/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2594 Acc: 0.9783\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3093 Acc: 0.9431\n",
      "\n",
      "Epoch 45/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2554 Acc: 0.9804\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.03it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3192 Acc: 0.9350\n",
      "\n",
      "Epoch 46/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2560 Acc: 0.9788\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.03it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3183 Acc: 0.9364\n",
      "\n",
      "Epoch 47/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:33<00:00,  1.33it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2575 Acc: 0.9799\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3134 Acc: 0.9392\n",
      "\n",
      "Epoch 48/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.32it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2542 Acc: 0.9809\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3136 Acc: 0.9403\n",
      "\n",
      "Epoch 49/49\n",
      "----------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [00:34<00:00,  1.31it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train Loss: 0.2543 Acc: 0.9804\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:55<00:00,  5.07it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val Loss: 0.3099 Acc: 0.9403\n",
      "\n",
      "Training complete in 75m 4s\n",
      "Best Train Acc: 0.9809\n",
      "Best Val Acc: 0.9431\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "model_ft = train_model(model, criterion, optimizer, exp_lr_scheduler,num_epochs=no_epochs) \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "6ca7adcb",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:45:52.548203Z",
     "iopub.status.busy": "2023-03-25T16:45:52.547819Z",
     "iopub.status.idle": "2023-03-25T16:45:52.981347Z",
     "shell.execute_reply": "2023-03-25T16:45:52.980338Z"
    },
    "papermill": {
     "duration": 1.385349,
     "end_time": "2023-03-25T16:45:52.984381",
     "exception": false,
     "start_time": "2023-03-25T16:45:51.599032",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "\n",
    "# save the model\n",
    "torch.save(model_ft.state_dict(), \"Deit_Model.pth\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "89f057ce",
   "metadata": {
    "papermill": {
     "duration": 0.930984,
     "end_time": "2023-03-25T16:45:54.803747",
     "exception": false,
     "start_time": "2023-03-25T16:45:53.872763",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Graphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "4ff80e8d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:45:57.083113Z",
     "iopub.status.busy": "2023-03-25T16:45:57.082748Z",
     "iopub.status.idle": "2023-03-25T16:45:57.315327Z",
     "shell.execute_reply": "2023-03-25T16:45:57.314370Z"
    },
    "papermill": {
     "duration": 1.224125,
     "end_time": "2023-03-25T16:45:57.317399",
     "exception": false,
     "start_time": "2023-03-25T16:45:56.093274",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjcAAAGwCAYAAABVdURTAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAA9hAAAPYQGoP6dpAABp+UlEQVR4nO3dd3gUdeLH8ffupockhASS0ELoHSUUQUEUBUFUzoaeIgioqKCIdyoip3LcwakolgOPE7CeoKL+vBMLShEBCyUUQURagCQEAul9d35/TLKwpJDAJpvyeT3PPJmdndn97hjZT77VYhiGgYiIiEgdYfV0AURERETcSeFGRERE6hSFGxEREalTFG5ERESkTlG4ERERkTpF4UZERETqFIUbERERqVO8PF2A6uZwOEhISCAoKAiLxeLp4oiIiEgFGIZBRkYGTZs2xWotv26m3oWbhIQEWrRo4eliiIiIyHk4fPgwzZs3L/ecehdugoKCAPPmBAcHe7g0IiIiUhHp6em0aNHC+T1ennoXboqbooKDgxVuREREapmKdClRh2IRERGpUzwabr777juuu+46mjZtisVi4dNPPz3nNWvXriU2NhY/Pz9at27N66+/XvUFFRERkVrDo+EmKyuLHj168Nprr1Xo/AMHDjB8+HAGDBjA1q1befLJJ3nooYdYvnx5FZdUREREaguP9rkZNmwYw4YNq/D5r7/+Oi1btmTevHkAdOrUiU2bNvHCCy9w0003VVEpRUREpDapVX1uNm7cyJAhQ1yODR06lE2bNlFQUFDqNXl5eaSnp7tsIiIiUnfVqnCTlJRERESEy7GIiAgKCws5ceJEqdfMnj2bkJAQ56Y5bkREROq2WhVuoOQQMMMwSj1ebNq0aaSlpTm3w4cPV3kZRURExHNq1Tw3kZGRJCUluRxLTk7Gy8uLsLCwUq/x9fXF19e3OoonIiIiNUCtqrnp168fK1eudDn29ddf06tXL7y9vT1UKhEREalJPBpuMjMziYuLIy4uDjCHesfFxREfHw+YTUp33XWX8/yJEydy6NAhpk6dyu7du1m8eDGLFi3iT3/6kyeKLyIiIjWQR5ulNm3axBVXXOF8PHXqVADGjBnDm2++SWJiojPoAMTExLBixQoeeeQR/vnPf9K0aVNeeeUVDQMXERERJ4tR3CO3nkhPTyckJIS0tDStLSUiIlJLVOb7u1Z1KBYREZHqZ3cYZOYVkpVXiM1qwcdmxcfL3LyslgotZlmdFG5ERESqUIHdQVZeIRm5hc6AkJlXiI+XlQAfLwJ9bPj72Aj08cLfx4avl9UZFgzDID23kKS0XBLTcjiWnktiWq7zZ1JaLgV2B40CfYo2X8ICfQgN9CHMecyHAB8bWXl2MoveOzOvgMzcQjKKy1O0n5lbeMY5px9n59vL/HwWC2bYOSPwNAn24/8evLS6bnEJCjciIiIV4HAYpOUUkJKVz6nsfFIy8zmZlc/JrDxOZhVwMiuPlKx8UrMLyHSGmQJyCxyVeh+b1UKAtxl4MnILySkoO1gU23c863w/VqV4WS04DAPHGR1aDAPyCh3kFTog7/QxT1K4ERGR82IYBkdO5fBLQhq7EtLZlZjOroR00nIKCPD1IsDHVqJmIsDHRoCvjSA/byKD/YgM8SMqxI/IYD/CGvhis1ZP80Z+oYOTWfmkZOUVBZR8TmXlk5pTQGp2AanZ5v6p7ALSivbTcgou6Evbz9tKA19vGvia96XA7iA73052vlkzkldohiC7wyAjz6xJKdYwwLxfUSHmPYsM9icyxJfIEH+8bRZn+VOKPktKVj4nM4tCWFY+Ofl2Gvh60cDPi0BfL4J8vZyPGxTtB/p6EeRnbg3OeD7I15tAXxsN/Lzw9bIBUGh3kG93kF9obnmF5uOComOepnAjIiLnlJlXyMETWexOPB1idiWmk5FbWOr5WeU0Y5TFy2ohItiPiGBfokL8aRzki7fNgtVi9umwWXHuWy1gs1iwWi0U2g3sDgcFDoNCu4NCh0Gh3Sj6aT5OP6PG5WRmvktwqKxgPy+XZqBGgd4uzUGNAr0J8vN2BoSgokDhbSt/9pVCu4OcAjvZ+XayipqCGvh6ERnih5+37bzLWxW8bFa8bFYCfDxdktIp3IiIlCM7v5DdiRnsSkwnMTWHdhEN6N68ITFhgVirqZahsgrsDo5n5HEsPZfkjDzSsgvw97GV+Ev97C/d7PxCDp7I5mBKFgdOZHHwRFbRfjYnMvNKfS9vm4X2EUF0aRpM56hgOjcNoXGQL9n5heTk28nKt5OTX0hWnp3sAjvZRV/aaTkFZj+S9FyOpeWSnJFLocPgaGoOR1NzgNQqv082q4XQAJ+icGKGlYYBPjT09yY0wIeQAPNnwwBvQgO8CfE3988VUs6Xl81KkM1KkJ8mpb1QCjciIkVOZObxS8LpWolfEtI4cCKr1KaIIF8vujUPoXvzhvRoHkL3Fg1pGuLnMmqkwO7gWHouCam5JBR9aSek5pCYlgtQ1FRjNlGYTThFzTi+Nvx9vPCyWiiwO7CfWRPhcFBQXFNhN8jOLyQ5PY/kojBzPMPs91EZft5W/LxtpGYXlHteWKAP7SIa0DkqhM5Ng+nSNJg2jRvg43XhX/aFdgfHM/PMzrJpZmfZE5l52A0DwzCbahxF+w7DKHpsNo1ZrRa8rRZsViveNgs2q8WsWbBa8LJZ8LJaCPLzPqO2xexsG+znXWMDqlwYzXMjIrXSqax8vv/9BGk55X8hG0BBoVndX1zVX9zH4cz9pDSzlqM0jYN86dI0mKgQf347lsHOo2nO/hFnCm/gQ6eoYLLz7SSkmiNbHB76F9bbZqFxA18aB/sRGuBNboHdZfRLZl5hqR1dGwZ40yoskJjwQFqFBdIqPICY8ECiwwIJ8VeNgniO5rkRkRrnWHouq39N5ttfk9l7LIMuTUO4pHUj+rUJo03jBhWaJ+PgiSy+2X2MlbuOsenQKexuTg4WC8SEBdK5aXBRzUQInaKCaBLk53Jeod3Bb8cy2X4klW1H0th+JJVfkzI4kZnPur0nXM71tlmICvGnaUM/mjb0p1lDf6JC/PGyWsjKPytsndV0Y3cYZs1DcS2E1YK3zVpUM2HB22rFz9scdts4yJeIYD+aBPnSJMiX0ACfc9ZKnDlEOTvfTkSwLw1raicKkUpQzY2IVAmHw2BnQhrf7k5m1a/J7DiaVua54Q18uaR1Iy5pHUa/NmG0Dg/EYrHgcBhsPZzqDDS/J2e6XNcxMohWYYHnLIuXzWKO1PE93fQTcMa8IoG+NhoG+NAhIohA3/P7my+3wM6uxHT2JGUQ4u9NVIgfzRr6E97AV00fIm5Qme9vhRsRKVN+oYMfD6Tw1S9JfLMrmdScfMIb+BZtPq77QeZ+Wk4Bq3Yns2pPMsfPaOaxWKBH84YM7tiErs1D2HEkjR/2p7D50KkSTTxNgnzp1iyEbUdSOZF5uv+Il9VC39aNuLpTBIM7RdCiUUC13QsR8SyFm3Io3Eh9UjzpWJCfF14VHOGRnV/I2j3H+eqXJL79NbnMob4VEehjY0C7xgzu1IRBHZrQOMi3xDm5BXa2HU5l4/4Uftifwpb4VJd5MoJ8vRjUsQlXFb2G+n2I1E/qcyNST2TnFxaNwjFH45w5Iich1ZyuvcBuYLWYTT+nJwDzIzLE3/k4vIEPW+JT+fqXJNbtPeFSkxLewIerO0cwpHMkrRsHciIznxOZeeaWccZ+Zh4nMvOxWnAGmj4xjZyTfpXFz9tG39Zh9G0dBphhZ2t8Kr8kpNExMpg+MY3cMhpHROoP1dyI1CJJabn8eCCFHw+c5KcDJ0v0QXGXlo0CGNolgqFdIrm4ZWi1zRorIlIW1dyIeMjhk9mk5xbg7RzdYi0a7XJ637v42DlW0jUMg8Mnc/jhQAo/FYWZ+JPZJc5r4OtFs4anR+MUj8hpWnSscZDZD+ZYWh6JaTkkpZuL7SWl5Tr3j6XnEh0WyNAukQzpEkHHyKAat8qviEhFKdyIXKC9xzL4fEciK3Yk8tuxitekOFfS9bLi62WuqOvtdXpl3ZTMfJLSc12usVqgS9MQ+sY0ok9MI2KjQwlrULIfy9maBNloEuRHt+Yhlf58IiK1jcKNyHkoK9B428zp3AsdhsvMsgUOR4lZbs9cSTejjPfxtlno3ryhS5jR1OwiIuVTuBGpoN+TM/jf9tIDzYB2jRneLYqrO0eUOZrH4TBDTqHdcK6cW7ySbvHKus7jdgcB3ja6N2+Iv0/NWjBPRKSmU7gRKYfDYbB6TzL/XrefH/afdB6vaKA5k9Vqwddq4zzniBOR+qQgF478DEGREN7O06WpdfTPrEgpcgvsfLL1KG+s28++41mAOYHcwPaVCzQiIhWWcwr2roRf/we/fwv5RTXE0ZdB7/HQcQR4eXB5DMOAwlwwSq5JVpIFfDw3yabCjcgZUjLzeOeHQ7yz8ZBzZeUgXy/+2LclY/q3omlDfw+XUETKlZsOybvg2E449ou5Jf8KAY2g1aXQagC0ugxCmlf+tQvzIP0oZJ0o2o6bW3bK6f2sFMjPgJAWZo1LWDvzZ3g785j1rGbm1MOwZwX8+jkcWg+OMybNDGxsvvah782tQQT0vAtix55f+c/mcED6ETi539xS4837l5cOeRlF+xmQl1b0M8O1fOVpEAl/2nPhZTxPmudGBNh3PJNF3x9g+eYjzgnsmjX05+5LWzGqdwt14hWpDg4HnNwHidsgabv5hWvzAW9/8A4o5WcAWL3Ma5J2moEm9VDF3qth9Omg0+pSaNjSPG4YkJkMKXvhxG9w4vei/b3ma1eo1qIMNl8IawNhbSG4GcRvMD/rmRp3go7XQsfhEHUxZCTA5rdgy1uQecw8x2KF9sPM2pzWV4C1jEkuC/Mg+yTknISMpKIQc+B0mDl1EOx5pV97oaog3Gj5hXIo3IhhGBxKyWZL/Ck2HzrFlvhUfk1Kd45m6t48hAkDWjO8a2SFlyyQeqi4ij4v0/xLPS/TbEYoyDG/KBu1LvlXurvlZUDKPgiKMv/KL+tLriyGYX75pR6EtKPgKDC/OLGYPy2WMx5bzJ+FuebnPPtzO39mmJ87sAk0aFz0M+KM/SYQEAYOOxzfDYnbzSCTuB2SdkBB1oXfl6CmENkVIrpARFdo3BEyEuHgOjj4PSTEgWF3vaZhSwgIh5TfzZqLsnj5F32Wxub5gY0hMLxoK9r3DjCDw4m9RcHodzOA2fNLvp7FCi0uMcNMh+Fm+CmNvcBsrvp5kfk5ijVqDTEDITftdJDJPmU2cVXkXlq9IbSV+TqhrcA/FHyDwC/Y/OkbbG7Fj30amIHyXCwWM4S6kcJNORRu6p/cAjvbj6Sx+ZAZZrbGn3I2ORWzWGBwxwjuGRBDn5hGmsCuqjkclf8iPt/32f1/5hdNgwjzi7VBpLkfEFZ2GXJOwalD5l/qpw6Z16cegvSEour5dPPL/OwvyDN5+UHjDtCkC0R0hiadzP2gyKKgcAEK8+HnN2DtPyA31Txm8zFrAxq2MJs/QlqYTRcNW4B/I7PsqUWf5czPll/WRARVyGI1t9KaOLz8zVAS1cO8f4YDCrLN0FiQc8Z+NuRnm2ErNNoMMRFFgSagUfnvn5cB8T+aIeHQeji6xfW/pcVqhp3iJqWwtqebmM73v5/Dbjb7pPxubqnxZlnbX2MGospI/hU2LYZt75cfxAAsNjOwBDaGRjFmiDlzC2le9SHcTRRuyqFwU/clpOYU1cicYsuhU/ySkE6hw/XX3MdmpVvzEHq2bEhsdCg9o0NpEuTnoRKfp/ws8y9dn0DzHy//Rh7twFchuWmw4jHY+ZFZ3kYxEBpT9LPV6f3AxhceAA7/BF88BglbS3/eYisKOxHmZvM2v3BOHTL7GFSGdyD4NjD/qrX5mAGiMKf0c/1DzZDTehBcdHvl+k4YBuz6FL55xnwPAN8QM6BcSHNJUJQZhrx8zdcxjKLXM0o+9vIzP2fx5y3+a/7Mx/YCyEqGzONFP4u2rGSzdoGi/x/9QiCyuxlkonqY+2FtwVbN3UHzMuHwj2ZgCmtrful7nXtyTI/Lz4JfPjF/b/0bmaHOvxEEhJ7+N8E3uHr+kKgGCjflULipW/ILHexKTDfDTFGgSUzLLXFekyBfYqNDiY0O5eKWoXRtFnzOBR1rrNR4+OnfZht87llfwl5+Z/wjV/QPXGA4NOsFba6A4KaeKTPAoY3w8b2QFn/uc70DzZDTfqjZgTK0VcXfJz0Rvnkati8zH/sGQ7shZnV9ZrLZ9yD7xLlfJ7CJWSPQMNr8GdrKDCJ+IeATdPrL3Cew5F++DrsZPpJ3wbFdkPyL+fPkvrNCiAXaXAkX32n2syjvCzX+R/j6KTjyk/m4QQRcMR0uugMwzJqZtCNFW/wZ+0fMzq/BTc/4PK3MrWG0WbPj5uaDctkLzfvvKDRrmlRLKhWkcFMOhZu6YfOhk7y26nc27EtxWcEawGa10DkqmJ4tG9IzOpSeLUNpHupfu5uaDAPif4AfF8Du/57+ggxsDFjML+6KjGJo3NHsgNjmCoi+1PyCrmr2QrP5ZN0LZrkbRsMNr5nB4NQBMwScLPp56qD5ZcxZ/yy1vgJix0CHa8seCluQCz/8E76bW9TXwGKGhsF/MWtoXMpUYI5syUgqqlVIMpt6GrYsCgAtzdDibgU5cHwPJMbBjo9c+074h0L3UWaZI7udPp6yz6yp2f2Z+dg7APo/BP0nV89/P5EaQuGmHAo3tduW+FO8tPI31u09/Zd3wwBvYluGOoNMjxYhBPjUkVkOCvPMaucf5ruOqogZCJc8YNZIWG1FazlkmCEn51RRx8Kin+lHizpRbnGtNbB6Q4u+ZtBpc4XZObIg26zqPnMrOGM/KNLs9FjRPgIn95u1NUd+Nh/3uB2GPWd2TizvM6fGm59367uwf/Xp5wLC4aI/Qs8xEN7WPGYY5lDar5483VTTvA8M+wc061mxcnrKyf2w9T2I+485KqZY1EVmyEnZZ/atKe7oe/GdMOhJCI7yWJFFPEXhphwKN7VT3OFUXlr5G2t/Ow6YE+rdHNuccZfF0K5Jg9pdK1OarBTzS+3nN8x+CmA2OXW/FfpONDsiVlbOKTjwHexbDftWVXzI7NksVnP4bOcboON1EBRR8hzDMDs7rvizOYLGNwRGvAjdbq78+506CFveMYNOZtLp49GXQfdb4JdPTwegoCi4eiZ0u6V2NXc47OZ/l61vw68rzDBzprZXm58rorNnyidSAyjclEPhpnbZfiSVed/sZdWv5he8zWrhpp7NmHRFO1qG1fDOs+frt6/h0/tP9wsJioLeEyD2bggMc9/7nNxvhpx9q83mkcJ8s0OyT6DZZORdvF+0efubw3UT4854EQu07GcGnU7XQUgzM0T97xGzxgmgZX+4caHZt+NC2Ath71fmnB+/r3SthbL5QL9JMODR2t9Uk5UCOz4w+wx5+cPlfzb75YjUcwo35VC4qR12Hk1j3je/8c1uM9RYLXBjz+ZMvrIt0WFV0BeiJijMg5VPm/1qwJzMa+CfzOBgq0GTCJ46CLs+g13/B0c3uT7XvI/ZsTX9iDka6Yon4bJH3D/UNO2IWZOz+39m89Tgv5gjXESkzlK4KYfCTc1mGAavr93Pc1/9imGYoWbkRc2YPLgdMeF1NNSA2cn0o/FwbIf5uO9EuOpZ8K7hw9NTD5sdnHf9Hxz+4fTx0Bi4aRE0j/Vc2USkTlG4KYfCTc2VW2Dn8eXb+b84s2Pltd2jmHp1e9o0rgHNDOkJ5lTsjTuV3sfkfBmGOaT7iyfMeVECwmDkAnMIdG2TnmDWpOSmmp2da3vzkIjUKJX5/q4jQ0qktktKy+Xedzax/UgaXlYLT1/fhdGXRHumMDmp5sRvRzebM5cmbDGnbi8W2R3aXQ1tr4Lmvc+/ySjnFHz20Okhvq0HwR/+ZY5Iqo2Cm0Lfez1dChERhRvxvC3xp7jvnc0cz8gjNMCb+XfE0q+NGzvOlsfhMCdYO7jeDDMJW8yp0c9msZoTuKXGm+vgJG2HdXPNCeJaDzKDTturzA61FXFoAyy/x+ybYvUy+4z0m1xnZhIVEfEkhRvxqI82H+HJj3eQb3fQISKIN8b0okWjKh4FlZ5oDh3et9r8mXW85DmhraBZrLk17QlR3c0RQ5nHYd+38Ps38Pu35rwyuz87XfvSuJM5CZzFZg5FttrM/TN/FuSYU+gbDrMT7E2Lav58LCIitYjH+9zMnz+f559/nsTERLp06cK8efMYMGBAmef/85//5LXXXuPgwYO0bNmS6dOnc9ddd1X4/dTnpmawOwzmfLGbf687AMCQzhG8OOoiGvhWMG8bhhkwUvaZ0+GXtvk0MGtC8rPMmpLiYc/Hd7u+lneAOVtviz5mkGnW89wL74E5N0lCXFHQ+cYcOVSZ9X16/BGGP2euxSMiIuWqNX1uli1bxpQpU5g/fz6XXnop//rXvxg2bBi7du2iZcuWJc5fsGAB06ZN49///je9e/fmp59+4p577iE0NJTrrrvOA59AzkdaTgEPvb/VOSHfQ1e2ZcpV7bFaKzjpWnoi/G8K/PZl+edZrGZwyM8+a1I0CzS9uGhm3ivN4ctlTelfHqvNHA3UPBYGPW7OBnzwe7NDrcNurjJsGKf3z/wZdRG0u6ry7ykiIufk0Zqbvn370rNnTxYsWOA81qlTJ0aOHMns2bNLnN+/f38uvfRSnn/+eeexKVOmsGnTJr7//vsKvadqbjzr8Mlsxiz+if0nsvDztjL3lou4tnsFp5I3DHOa+q+mmQtGWr3NUUUFOeZj55YK9nzXa0Nanl5mIObyitXMiIhIjVEram7y8/PZvHkzTzzxhMvxIUOGsGHDhlKvycvLw8/Pdd4Pf39/fvrpJwoKCvD2LjlqJS8vj7y8POfj9PR0N5RezkdGbgHj3/qZpBMpNA0OYeGY3nRtFlKxi9OOwH+nmDPTglnzcsP8sqejL8g9HXa8fMzFGmvTdPwiInLePDY048SJE9jtdiIiXOcMiYiIICkpqdRrhg4dyhtvvMHmzZsxDINNmzaxePFiCgoKOHHiRKnXzJ49m5CQEOfWosUFTgEv58XhMPjHO5/x55PPsstvHOu8H6Trlr/Ani/MZqOyGIY53f78fmawsfnCVc/A+G/KX2fH28+cj6Zxe7NzsIKNiEi94fHRUmcveGgYRpmLIM6YMYOkpCQuueQSDMMgIiKCsWPH8txzz2GzlT69+7Rp05g6darzcXp6ugJOdcs8zrZ3HueZpE/wspkdbm1ZSbD5TXPz8jOHU7e/xmxmCm5qXpcab84DU7woYvPeZm1N4/ae+BQiIlJLeCzchIeHY7PZStTSJCcnl6jNKebv78/ixYv517/+xbFjx4iKimLhwoUEBQURHh5e6jW+vr74+vq6vfxSAfnZ8MN8Cr57kYsLs8ACiZFXEDXyr5B5DPZ8aXYKTjts/izuIBzVw+zku+19c0VpLz+4cgZccr/71ygSEZE6x2PhxsfHh9jYWFauXMkf/vAH5/GVK1dyww03lHutt7c3zZs3B2Dp0qWMGDECqyY/qzkcDnNF41V/hfSjeAPbHTHs7PJn/jjqjqKTupmT3g1/Ho79cjrcHNkEidvMDcwVp69/zVwcUUREpAI82iw1depURo8eTa9evejXrx8LFy4kPj6eiRMnAmaT0tGjR3n77bcB+O233/jpp5/o27cvp06d4sUXX2Tnzp289dZbnvwYcqb9a+DrpyDJXAAykXBm548iq90NLLylT8nzLRaI7GpuA/8Emcmw92uI32hOoNdzrGbtFRGRSvFouBk1ahQpKSnMnDmTxMREunbtyooVK4iONtcUSkxMJD4+3nm+3W5n7ty57NmzB29vb6644go2bNhAq1atPPQJxMkw4IvH4KeF5kPfYN7yuonZKZfTskkjPr69J7aKzGPToAlcfKe5iYiInAePz1Bc3TTPTRUwDLO2ZuNrgAWjzz3MOHUt7+7IIsTfm88mXUp0WKCnSykiIrVYZb6/Vd8vF27tP4qCDXD9KywMnMi7O7KwWS0suKOngo2IiFQrhRu5MBtehTVFs0lfM4dVAUOZ8+WvADx9XWf6ty19FJuIiEhVUbiR8/fzIrM5CuDKp/i99Z089H4chgG392nJ6EuiPVs+ERGplxRu5PxsWwqfP2ruX/YIjksf5dEPt5OZV0ifmEY8e32XMidjFBERqUoKN1J5uz6DT+8HDOhzLwx+mv/uSGTb4VQCfWy8dvvF+HjpV0tERDxD30BSOXu/gY/GgeGAi+6Aa/5BToGDf3xh9rN54Iq2NAn2O8eLiIiIVB2FG6m4g9/DsjvAUQCdR8L1r4LVyqLv95OQlkuzhv6MvyzG06UUEZF6TuFGKubIZvjPKCjMhXZD4cZ/g9VGckYu89fsA+Cxazrg5621n0RExLMUbuTcck7B0tvNRSxjBsKtb4GXDwBzv/qN7Hw7F7VoyPU9mnq4oCIiIgo3UhFfP2Wu4h3WDm57H7z9AdiVkM4Hmw8DMGNEZ42OEhGRGkHhRsq3bzVsfRewwA2vgW8DAAzDYNbnuzAMGNE9itjoUM+WU0REpIjCjZQtPwv++5C53+ceaHmJ86lvdyezYV8KPl5WHr+mo4cKKCIiUpLCjZRt1SxIjYeQFjD4L87DBXYHf1+xG4Dxl8XQolGAp0ooIiJSgsKNlO7wz/DDAnN/xDzwDXI+9d4Ph9h/IouwQB8eGNTGM+UTEREpg8KNlFSYB59NAgzocTu0u8r5VFp2AfO+3QvA1CHtCfLz9lAhRURESqdwIyWtmwvHf4XAxjD07y5PvbJqL6nZBbSPaMCoXi08VEAREZGyKdyIq6SdZrgBGP48BDRyPnXgRBZvbzwIwFPXdsbLpl8fERGpefTtJKfZC83mKEchdBxhLrFwhjlf7KbAbjCoQ2MGtm/smTKKiIicg8KNnPbjAkjYCr4hMPwFOGNSvo37Uvjql2PYrBamD+/kwUKKiIiUT+FGTCn7YNXfzP2hsyA4yvmUw2HwtxW7ALi9TwvaRQSV9goiIiI1gsKNgGHAfx+Gwhxz7aiLR7s8/cnWo+w8mk6QrxePXNXeQ4UUERGpGIUbgS1vwcF14OUP173i0hyVnV/I81/tAeDBK9sS1sDXU6UUERGpEIWb+u7UQfh6hrk/eAY0inF5+t/fHSApPZfmof6M7d+q2osnIiJSWQo39VlhPnw0DvLSoXkf6DvR5elj6bm8vnYfAI9f0xE/b5snSikiIlIpCjf12aqZcHQz+IXAzYvA6hpe5n69h5wCOxe3bMiI7lFlvIiIiEjNonBTX/32FWx41dy/YT40bOny9K6EdD7cfAQwJ+yznNEPR0REpCZTuKmP0o7CJ0VNUH3ug04jXJ42DHPot2HAiO5RxEaHeqCQIiIi50fhpr6xF8LH90DOSYjqAUP+WuKU1XuSWf97Cj5eVh6/pqMHCikiInL+FG7qm++eg0PrwacB3LwEvFyHdhfYHfzt890A3H1pK1o0CvBEKUVERM6bwk19sn8trH3O3L/uZQhrU+KU93+KZ9/xLBoF+vDgFW2ruYAiIiIXTuGmvsg8bjZHYZgzEHe7ucQpaTkFzPtmLwCPXNWOYD/vai6kiIjIhVO4qQ8cDvjkXsg8Bo07wrDnSj1t/urfOZmVT9smDbi9T8tSzxEREanpFG7qgw0vw75V5vIKt7wJPiX70Rw+mc2S9QcBeHJ4R7xs+tUQEZHaSd9gdV38j/Bt0Yio4c9Bk06lnjbny1/Jtzu4rG04V3RoUo0FFBERcS+Fm7osNx2WjwfDDt1uKbHad7HNh07x+fZELBZ4cngnTdgnIiK1msJNXfbr/yDtsDn78IiXXFb7LmYYBrM+3wXArbEt6Nw0uLpLKSIi4lYeDzfz588nJiYGPz8/YmNjWbduXbnnv/fee/To0YOAgACioqK4++67SUlJqabS1jIHvjN/dr0JfINKPeXHAyfZGp+Kv7eNR4e0r8bCiYiIVA2Phptly5YxZcoUpk+fztatWxkwYADDhg0jPj6+1PO///577rrrLsaPH88vv/zChx9+yM8//8yECROqueS1gGHAgaKg2GpAmae996N5r0de3IwmwX7VUTIREZEq5dFw8+KLLzJ+/HgmTJhAp06dmDdvHi1atGDBggWlnv/DDz/QqlUrHnroIWJiYrjsssu477772LRpUzWXvBY4dQDSj4DVG1peUuopxzPy+HJnIgB39NXQbxERqRs8Fm7y8/PZvHkzQ4YMcTk+ZMgQNmzYUOo1/fv358iRI6xYsQLDMDh27BgfffQR1157bZnvk5eXR3p6ustWLxTX2jTvBT6BpZ7ywabDFNgNLmrRkK7NQqqxcCIiIlXHY+HmxIkT2O12IiIiXI5HRESQlJRU6jX9+/fnvffeY9SoUfj4+BAZGUnDhg159dVXy3yf2bNnExIS4txatGjh1s9RYx0sv0nK7jD4T1GT1J2XRFdXqURERKqcxzsUnz3s2DCMMoci79q1i4ceeoi//OUvbN68mS+//JIDBw4wceLEMl9/2rRppKWlObfDhw+7tfw10pn9bWJKDzdrf0vmaGoOIf7ejOgeVY2FExERqVpennrj8PBwbDZbiVqa5OTkErU5xWbPns2ll17Kn//8ZwC6d+9OYGAgAwYMYNasWURFlfyS9vX1xdfXt8TxOi3ld8hMApsvNO9T6inv/mDW2twS2xw/b1t1lk5ERKRKeazmxsfHh9jYWFauXOlyfOXKlfTv37/Ua7Kzs7FaXYtss5lfzIZhVE1Ba6PiIeAt+oB3yRFQh09ms3pPMgB3qElKRETqGI82S02dOpU33niDxYsXs3v3bh555BHi4+OdzUzTpk3jrrvucp5/3XXX8fHHH7NgwQL279/P+vXreeihh+jTpw9Nmzb11Meoec7R3+b9n+IxDLisbTgx4aV3NhYREamtPNYsBTBq1ChSUlKYOXMmiYmJdO3alRUrVhAdbdYmJCYmusx5M3bsWDIyMnjttdd49NFHadiwIVdeeSX/+Mc/PPURah7DgIPfm/ul9LfJL3TwwSaz39Gdl2j4t4iI1D0Wo56156SnpxMSEkJaWhrBwXVwqYHk3TD/EnMF8CcOgZdrf6P/bktg8vtbaRLky/onrsRbq3+LiEgtUJnvb32z1TXFo6Ra9i0RbADe/eEQALf1aalgIyIidZK+3eqag0WdiUvpb7P3WAY/HjiJzWrh9j71ZL4fERGpdxRu6hKH44z+NgNLPF28jtTgjk2ICvGvzpKJiIhUG4WbuuTYTsg5BT4NoOnFLk9l5xeyfPMRQDMSi4hI3aZwU5cUDwFv2Q9s3i5PfRaXQEZeIdFhAVzWNtwDhRMREakeCjd1SRlLLhiGwbs/mh2J7+jbEqu19OUtRERE6gKFm7rCYYdDRaupn9WZeNuRNHYeTcfHy8rNsepILCIidZvCTV2RuA3y0sA3BKJ6uDz1XtHw72u7RdEo0McTpRMREak2Cjd1RXF/m+j+YD29EGZadgH/3Z4AaEZiERGpHxRu6ooy+tt8tOUIuQUOOkYG0bNlqAcKJiIiUr0UbuoCewHEbzT3z+hvYxgG7xV1JL7zkmgsFnUkFhGRuk/hpi5IiIP8TPAPhYiuzsM/7D/J/uNZBPrYGHlxM8+VT0REpBop3NQFxUsuRF8K1tP/ST8sWv37+oua0cDXowvAi4iIVBuFm7rA2d/m9JIL6bkFrNiZCMCtvZp7olQiIiIeoXBT2xXmQ/wP5v4Z/W0+355IboGDtk0acFGLhp4pm4iIiAco3NR2RzdBYQ4EhEOTTs7DxU1St8Q2V0diERGpVxRuarszh4AXhZjfkzPYEp+KzWrhDz3VkVhEROoXhZvarnjyvjOapD4sWv37ig6NaRLk54lSiYiIeIzCTW1WkAuHfzL3izoTF9odfLzlKIDWkRIRkXpJ4aY2O/IT2POgQSSEtQVg7W/HOZ6RR1igD1d2bOLhAoqIiFQ/hZvarJT+Nh9uMpukRl7cDB8v/ecVEZH6R99+tdlZ/W1SMvP4ZvcxAG7R3DYiIlJPKdzUVvnZcGSTuV+0WOancQkUOgy6Nw+hY2SwBwsnIiLiOQo3tdXhH8BRAMHNITQGwzBc5rYRERGprxRuaqviWptWl4LFws6j6fyalIGPl5Xre2huGxERqb8UbmqrNLOWhkZtAPhws/l4aJdIQgK8PVUqERERj1O4qa3SzUUxCW5KboGd/4tLANQkJSIionBTW6WbYYbgKFbuOkZaTgFRIX5c2jbcs+USERHxMIWb2irdnIWY4GbO5RZujm2OzapFMkVEpH5TuKmN8rMhNxWARKMR6/YeB8xwIyIiUt8p3NRGGUX9bXwasHxnGoYBfWIaER0W6NlyiYiI1AAKN7VRUZOUERTFR0WLZN7aS4tkioiIgMJN7VTUmTjdpwkHU7IJ9LExvFukhwslIiJSMyjc1EZF4ea3nCAAru0eRYCPlydLJCIiUmMo3NRGReFmyyl/QE1SIiIiZ1K4qY2Kws3hwoa0Dg8kNjrUwwUSERGpOTwebubPn09MTAx+fn7Exsaybt26Ms8dO3YsFoulxNalS5dqLHENkGGGm0SjETdc1AyLRXPbiIiIFPNouFm2bBlTpkxh+vTpbN26lQEDBjBs2DDi4+NLPf/ll18mMTHRuR0+fJhGjRpxyy23VHPJPcsoqrlJMhoxuFMTD5dGRESkZvFouHnxxRcZP348EyZMoFOnTsybN48WLVqwYMGCUs8PCQkhMjLSuW3atIlTp05x9913V3PJPcheAJnJABQERtE5KtjDBRIREalZPBZu8vPz2bx5M0OGDHE5PmTIEDZs2FCh11i0aBFXXXUV0dHRZZ6Tl5dHenq6y1arZSRhwSDP8OKi9q2xarkFERERFx4LNydOnMButxMREeFyPCIigqSkpHNen5iYyBdffMGECRPKPW/27NmEhIQ4txYtavnIoqImqWNGKIM6aW4bERGRs3m8Q/HZnWENw6hQB9k333yThg0bMnLkyHLPmzZtGmlpac7t8OHDF1JcjzuesB+AYzTisnZaAVxERORsHpv5LTw8HJvNVqKWJjk5uURtztkMw2Dx4sWMHj0aHx+fcs/19fXF19f3gstbUxzcv5fGQH5AJMF+3p4ujoiISI3jsZobHx8fYmNjWblypcvxlStX0r9//3KvXbt2Lb///jvjx4+vyiLWSCmJBwFo0LilZwsiIiJSQ3l0zv6pU6cyevRoevXqRb9+/Vi4cCHx8fFMnDgRMJuUjh49yttvv+1y3aJFi+jbty9du3b1RLE9JiffjiPtKFghqkUbTxdHRESkRvJouBk1ahQpKSnMnDmTxMREunbtyooVK5yjnxITE0vMeZOWlsby5ct5+eWXPVFkj9q4/wRNOAlA42YxHi6NiIhIzWQxDMPwdCGqU3p6OiEhIaSlpREcXLvmiJnx6U7u23oDzS0nYPw30KK3p4skIiJSLSrz/e3x0VJSMYZhsObXJCI4ZR4IjvJsgURERGqoSoebVq1aMXPmzDKXSJCqse94Jrmpx/C22DEsVmhQ/ogyERGR+qrS4ebRRx/l//7v/2jdujVXX301S5cuJS8vryrKJmdY/etxIixmfxtLgwiwaRi4iIhIaSodbiZPnszmzZvZvHkznTt35qGHHiIqKopJkyaxZcuWqiijAKt+TSaqKNwQpCYpERGRspx3n5sePXrw8ssvc/ToUZ5++mneeOMNevfuTY8ePVi8eDH1rJ9ylcrILeDngyeJLA43wU09WyAREZEa7LyHghcUFPDJJ5+wZMkSVq5cySWXXML48eNJSEhg+vTpfPPNN/znP/9xZ1nrrfW/n6DQYdAxKBMKgOBmni6SiIhIjVXpcLNlyxaWLFnC+++/j81mY/To0bz00kt07NjRec6QIUMYOHCgWwtan63+9TgAXYMy4SSquRERESlHpcNN7969ufrqq1mwYAEjR47E27tkx9bOnTtz2223uaWA9Z1hGKzekwxAtHeqeVDhRkREpEyVDjf79+93ziBclsDAQJYsWXLehZLTfklIJzkjD39vG8EFZg2Owo2IiEjZKt2hODk5mR9//LHE8R9//JFNmza5pVBy2pqiWptL24RhzUg0DyrciIiIlKnS4ebBBx/k8OHDJY4fPXqUBx980C2FktNW7zFra4a29YOCbPOghoKLiIiUqdLhZteuXfTs2bPE8Ysvvphdu3a5pVBiOpWVz9Z4c7mFgZEF5kH/RuDt78FSiYiI1GyVDje+vr4cO3asxPHExES8vDy6yHid893e4zgM6BgZRIRRPMeNhoGLiIiUp9Lh5uqrr2batGmkpaU5j6WmpvLkk09y9dVXu7Vw9d3qX83+NoM6NIH0o+ZBLZgpIiJSrkpXtcydO5eBAwcSHR3NxRdfDEBcXBwRERG88847bi9gfWV3GKz9zexvc0WHxnAowXxCnYlFRETKVelw06xZM7Zv3857773Htm3b8Pf35+677+b2228vdc4bOT/bjqRyKruAID8vekaHws7icKNmKRERkfKcVyeZwMBA7r33XneXRc6wpqhJamC7xnjbrJCumhsREZGKOO8ewLt27SI+Pp78/HyX49dff/0FF0pgVdH8Nld0bGIeKA43GgYuIiJSrvOaofgPf/gDO3bswGKxOFf/tlgsANjtdveWsB5KTs9l59F0AC5v39g8mK5mKRERkYqo9Giphx9+mJiYGI4dO0ZAQAC//PIL3333Hb169WLNmjVVUMT6Z01RR+LuzUNoHOQL+VmQm2o+qWYpERGRclW65mbjxo2sWrWKxo0bY7VasVqtXHbZZcyePZuHHnqIrVu3VkU565XiUVKDnLU2Rcsu+DQAv2APlUpERKR2qHTNjd1up0GDBgCEh4eTkGA2l0RHR7Nnzx73lq4eMgyDTQfNCfv6tw03DzrnuFGtjYiIyLlUuuama9eubN++ndatW9O3b1+ee+45fHx8WLhwIa1bt66KMtYrR07lcCw9Dy+rhR7NG5oHtWCmiIhIhVU63Dz11FNkZWUBMGvWLEaMGMGAAQMICwtj2bJlbi9gfbOlaC2pLk2D8fexmQeLa26CFG5ERETOpdLhZujQoc791q1bs2vXLk6ePEloaKhzxJScv00HzXATG93o9EHNcSMiIlJhlepzU1hYiJeXFzt37nQ53qhRIwUbN9l8qDjchJ4+mK5mKRERkYqqVLjx8vIiOjpac9lUkcy8Qn5NMue36dXqzHBT3KFYc9yIiIicS6VHSz311FNMmzaNkydPVkV56rW4+FQcBjRr6E9EsN/pJ5zNUpqdWERE5Fwq3efmlVde4ffff6dp06ZER0cTGBjo8vyWLVvcVrj6ZtMhMzC61NoU5kOWOe+Nam5ERETOrdLhZuTIkVVQDIEy+ttkJgEG2HwgIMwzBRMREalFKh1unn766aooR71ndxjExacCZ3cmPmPBTHXaFhEROadK97mRqvHbsQwy8goJ9LHRISLo9BPqTCwiIlIpla65sVqt5Q771kiq81PcJHVxy1C8bGdkTg0DFxERqZRKh5tPPvnE5XFBQQFbt27lrbfe4tlnn3Vbweqb4nDT88wmKdBIKRERkUqqdLi54YYbShy7+eab6dKlC8uWLWP8+PFuKVh9UxxuepUIN2qWEhERqQy39bnp27cv33zzjbterl5Jzsgl/mQ2Fgtc1LKh65NaNFNERKRS3BJucnJyePXVV2nevHmlr50/fz4xMTH4+fkRGxvLunXryj0/Ly+P6dOnEx0dja+vL23atGHx4sXnW/QaYUtRrU2HiCCC/bxdn3SOllK4ERERqYhKN0udvUCmYRhkZGQQEBDAu+++W6nXWrZsGVOmTGH+/Plceuml/Otf/2LYsGHs2rWLli1blnrNrbfeyrFjx1i0aBFt27YlOTmZwsLCyn6MGuX0YplnNUk5HKq5ERERqaRKh5uXXnrJJdxYrVYaN25M3759CQ0NLefKkl588UXGjx/PhAkTAJg3bx5fffUVCxYsYPbs2SXO//LLL1m7di379++nUSNz1exWrVpV9iPUOJvji/rbtDrr/mUdB0chWKzQIMIDJRMREal9Kh1uxo4d65Y3zs/PZ/PmzTzxxBMux4cMGcKGDRtKveazzz6jV69ePPfcc7zzzjsEBgZy/fXX89e//hV/f/9Sr8nLyyMvL8/5OD093S3ld5fcAjs7j6YBENuykeuTxZ2JG0SCrdL/qUREROqlSn9jLlmyhAYNGnDLLbe4HP/www/Jzs5mzJgxFXqdEydOYLfbiYhwrZGIiIggKSmp1Gv279/P999/j5+fH5988gknTpzggQce4OTJk2X2u5k9e3aNHqK+42gaBXaDxkG+tGh0VkDTMHAREZFKq3SH4jlz5hAeHl7ieJMmTfj73/9e6QKcPSGgYRhlThLocDiwWCy899579OnTh+HDh/Piiy/y5ptvkpOTU+o106ZNIy0tzbkdPny40mWsSs7+Ni1DS35u9bcRERGptErX3Bw6dIiYmJgSx6Ojo4mPj6/w64SHh2Oz2UrU0iQnJ5eozSkWFRVFs2bNCAkJcR7r1KkThmFw5MgR2rVrV+IaX19ffH19K1yu6uac3+bs/jagOW5ERETOQ6Vrbpo0acL27dtLHN+2bRthYRVftdrHx4fY2FhWrlzpcnzlypX079+/1GsuvfRSEhISyMzMdB777bffsFqt5zUM3dMMw2BLfBkzE4PropkiIiJSIZUON7fddhsPPfQQq1evxm63Y7fbWbVqFQ8//DC33XZbpV5r6tSpvPHGGyxevJjdu3fzyCOPEB8fz8SJEwGzSemuu+5ynv/HP/6RsLAw7r77bnbt2sV3333Hn//8Z8aNG1dmh+Ka7MCJLE5m5ePjZaVr05CSJzj73KjmRkREpKIq3Sw1a9YsDh06xODBg/HyMi93OBzcddddle5zM2rUKFJSUpg5cyaJiYl07dqVFStWEB0dDUBiYqJLU1eDBg1YuXIlkydPplevXoSFhXHrrbcya9asyn6MGmFTUZNUj+Yh+HiVkjOd4UZ9bkRERCrKYhiGcT4X7t27l7i4OPz9/enWrZszkNR06enphISEkJaWRnBwsEfL8sTy7Sz9+TATL2/DE8M6uj5pGPC3KCjMgYe2QqPWnimkiIhIDVCZ7+/znjylXbt2pXbglYor7kxcYmZigJxTZrABLb0gIiJSCZXuc3PzzTczZ86cEseff/75EnPfSNlSs/PZm2x2jC413BQPAw8IA2+/aiyZiIhI7VbpcLN27VquvfbaEsevueYavvvuO7cUqj7YGp8KQOvwQBoF+pQ8Qf1tREREzkulw01mZiY+PiW/jL29vWvc0gY12aZDJ4Eyam3g9Bw3apISERGplEqHm65du7Js2bISx5cuXUrnzp3dUqj6oNz+NgDpmp1YRETkfFS6Q/GMGTO46aab2LdvH1deeSUA3377Lf/5z3/46KOP3F7AuqjA7iDucCpQxszEoNmJRUREzlOlw83111/Pp59+yt///nc++ugj/P396dGjB6tWrfL40OraYndiOrkFDkL8vWkd3qD0k7RopoiIyHk5r6Hg1157rbNTcWpqKu+99x5Tpkxh27Zt2O12txawLnIulhkditVa+iKh6lAsIiJyfird56bYqlWruPPOO2natCmvvfYaw4cPZ9OmTe4sW521Of4c/W0AMrT0goiIyPmoVM3NkSNHePPNN1m8eDFZWVnceuutFBQUsHz5cnUmriDDMNh88BzhJi8TctPMfS2aKSIiUikVrrkZPnw4nTt3ZteuXbz66qskJCTw6quvVmXZ6qSEtFyS0nPxslro0bxh6ScVT+DnEwR+6sckIiJSGRWuufn666956KGHuP/++7XswgXYdNCc36ZL02D8fWyln6T+NiIiIuetwjU369atIyMjg169etG3b19ee+01jh8/XpVlq5O2FM1v07O8/jYaKSUiInLeKhxu+vXrx7///W8SExO57777WLp0Kc2aNcPhcLBy5UoyMjKqspx1gmEY/HjArLnpFd2o7BM1x42IiMh5q/RoqYCAAMaNG8f333/Pjh07ePTRR5kzZw5NmjTh+uuvr4oy1hkb96Xwa1IGPl5WLmldTrjJ0OzEIiIi5+u8h4IDdOjQgeeee44jR47w/vvvu6tMdda8b/cCcHvvFoQ18C37xOJmKY2UEhERqbQLCjfFbDYbI0eO5LPPPnPHy9VJG/el8NOBk/jYrEwc1Kb8k9UsJSIict7cEm7k3F7+9jcARvVuQVSIf/kna7SUiIjIeVO4qQY/7k/hh/0n8bZZuP9ctTaF+ZBVNApNNTciIiKVpnBTDV4u6mtza68WNG14jlqb4s7ENl8IKKfTsYiIiJRK4aaK/XzwJBv2peBts/DAFW3PfcGZc9xYylhUU0RERMqkcFPFXimqtbk5tgXNzlVrA1owU0RE5AIp3FShzYdOsW7vCbysFh44V1+bYupMLCIickEUbqrQy85am+a0aBRQsYs0x42IiMgFUbipIlvjT/Hdb8exWS08MKgCfW2KaY4bERGRC6JwU0WKa21uvLgZLcMqWGsDkK6lF0RERC6Ewk0ViDucypo9Zq3NpCsrUWsD6nMjIiJygRRuqkDxCKmRFzUjOiyw4hc67Fo0U0RE5AIp3LjZ9iOprPo1GauFytfaZCaDYQeLDRpEVE0BRURE6jiFGzc7s9YmJrwStTZweo6boEiw2txcMhERkfpB4caNdh5N45vd51lrAxoGLiIi4gYKN25UPELq+h5Nad24QeVfQJ2JRURELpjCjZv8kpDGyl3HsFhg0pXtzu9F0rX0goiIyIXy8nQB6opmDf2ZdEVbTmbn07bJedTagOuimSIiInJeFG7cpGGAD38a2uHCXkQ1NyIiIhdMzVI1iXPpBfW5EREROV8eDzfz588nJiYGPz8/YmNjWbduXZnnrlmzBovFUmL79ddfq7HEVcQwNIGfiIiIG3g03CxbtowpU6Ywffp0tm7dyoABAxg2bBjx8fHlXrdnzx4SExOdW7t259mBtybJOQWFuea+hoKLiIicN4+GmxdffJHx48czYcIEOnXqxLx582jRogULFiwo97omTZoQGRnp3Gy2OjDhXXGTVEA4ePl6tiwiIiK1mMfCTX5+Pps3b2bIkCEux4cMGcKGDRvKvfbiiy8mKiqKwYMHs3r16nLPzcvLIz093WWrkTTHjYiIiFt4LNycOHECu91ORITrGkoREREkJSWVek1UVBQLFy5k+fLlfPzxx3To0IHBgwfz3Xfflfk+s2fPJiQkxLm1aNHCrZ/DbTRSSkRExC08PhTcYrG4PDYMo8SxYh06dKBDh9PDrfv168fhw4d54YUXGDhwYKnXTJs2jalTpzofp6en18yAozluRERE3MJjNTfh4eHYbLYStTTJycklanPKc8kll7B3794yn/f19SU4ONhlq5HULCUiIuIWHgs3Pj4+xMbGsnLlSpfjK1eupH///hV+na1btxIVVQdqO5xz3KhZSkRE5EJ4tFlq6tSpjB49ml69etGvXz8WLlxIfHw8EydOBMwmpaNHj/L2228DMG/ePFq1akWXLl3Iz8/n3XffZfny5SxfvtyTH8M9iue40TBwERGRC+LRcDNq1ChSUlKYOXMmiYmJdO3alRUrVhAdHQ1AYmKiy5w3+fn5/OlPf+Lo0aP4+/vTpUsXPv/8c4YPH+6pj+A+6lAsIiLiFhbDMAxPF6I6paenExISQlpaWs3pf5OXAbObm/vTjoBvkGfLIyIiUsNU5vvb48svCJBe1CTlG6JgIyIicoEUbmoCZ2di9bcRERG5UAo3NYGGgYuIiLiNwk1NoHAjIiLiNgo3NUFGUbgJUrgRERG5UAo3NYFqbkRERNxG4aYm0OzEIiIibqNwUxOo5kZERMRtFG48rSAXslPMfYUbERGRC6Zw42nFa0p5+YF/qGfLIiIiUgco3HjamU1SFotnyyIiIlIHKNx4mnM1cDVJiYiIuIPCjac5R0op3IiIiLiDwo2naaSUiIiIWynceJrmuBEREXErhRtPSy/qc6MVwUVERNxC4cbT1CwlIiLiVgo3nmQvhMwkc1/NUiIiIm6hcONJmcfAcIDFBoGNPV0aERGROkHhxpOcc9xEgdXm2bKIiIjUEQo3nqQ5bkRERNxO4caT1JlYRETE7RRuPEk1NyIiIm6ncONJzjluFG5ERETcReHGk9QsJSIi4nYKN55U3CylFcFFRETcRuHGUwzj9FBw1dyIiIi4jcKNp2SngD3f3A/SulIiIiLuonDjKcVNUoFNwMvHs2URERGpQxRuPMXZmVi1NiIiIu6kcOMpznCjBTNFRETcSeHGUzQMXEREpEoo3HhKcbhRZ2IRERG3UrjxFOfSC2qWEhERcSeFG0/RHDciIiJVQuHGEwwD0lRzIyIiUhU8Hm7mz59PTEwMfn5+xMbGsm7dugpdt379ery8vLjooouqtoBVIS8dCrLMfQ0FFxERcSuPhptly5YxZcoUpk+fztatWxkwYADDhg0jPj6+3OvS0tK46667GDx4cDWV1M2KOxP7hYBPoGfLIiIiUsd4NNy8+OKLjB8/ngkTJtCpUyfmzZtHixYtWLBgQbnX3Xffffzxj3+kX79+1VRSN9McNyIiIlXGY+EmPz+fzZs3M2TIEJfjQ4YMYcOGDWVet2TJEvbt28fTTz9doffJy8sjPT3dZfM4DQMXERGpMh4LNydOnMButxMREeFyPCIigqSkpFKv2bt3L0888QTvvfceXl5eFXqf2bNnExIS4txatGhxwWW/YJrAT0REpMp4vEOxxWJxeWwYRoljAHa7nT/+8Y88++yztG/fvsKvP23aNNLS0pzb4cOHL7jMFyxDzVIiIiJVpWLVH1UgPDwcm81WopYmOTm5RG0OQEZGBps2bWLr1q1MmjQJAIfDgWEYeHl58fXXX3PllVeWuM7X1xdfX9+q+RDnSzU3IiIiVcZjNTc+Pj7ExsaycuVKl+MrV66kf//+Jc4PDg5mx44dxMXFObeJEyfSoUMH4uLi6Nu3b3UV/cIp3IiIiFQZj9XcAEydOpXRo0fTq1cv+vXrx8KFC4mPj2fixImA2aR09OhR3n77baxWK127dnW5vkmTJvj5+ZU4XuM5l15QuBEREXE3j4abUaNGkZKSwsyZM0lMTKRr166sWLGC6OhoABITE885502tU5ADOafMfYUbERERt7MYhmF4uhDVKT09nZCQENLS0ggODq7+AqTsg1d7gpc/TE+EUjpPi4iIiKvKfH97fLRUvXNmfxsFGxEREbdTuKlu6kwsIiJSpRRuqpvmuBEREalSCjfVzVlzo6UXREREqoLCTXXTopkiIiJVSuGmuqnPjYiISJVSuKluqUXz9ijciIiIVAmFm+qUcQyyT4DFCuEdPF0aERGROknhpjol7TB/NmoDPgGeLYuIiEgdpXBTnZK2mz8ju3m2HCIiInWYwk11Kq65UbgRERGpMgo31enYTvOnwo2IiEiVUbipLvlZcGKvua9wIyIiUmUUbqpL8m7AgMDG0CDC06URERGpsxRuqsuZ/W20GriIiEiVUbipLsXhJqKrZ8shIiJSxyncVBdnzU13z5ZDRESkjlO4qQ4OBxz7xdxXZ2IREZEqpXBTHU4dgIIssPlCWFtPl0ZERKRO8/J0AeqF4pmJIzqDTbdcRM6fw+EgPz/f08UQqRI+Pj5YrRde76Jv2uqQpMn7ROTC5efnc+DAARwOh6eLIlIlrFYrMTEx+Pj4XNDrKNxUB+dIKYUbETk/hmGQmJiIzWajRYsWbvnrVqQmcTgcJCQkkJiYSMuWLbFcwLQpCjfVQWtKicgFKiwsJDs7m6ZNmxIQEODp4ohUicaNG5OQkEBhYSHe3t7n/TqK/lUtKwUyEsz9iC6eLYuI1Fp2ux3ggqvrRWqy4t/v4t/386VwU9WOFdXahMaAX7BnyyIitd6FVNWL1HTu+v1WuKlqziYpzUwsIiJSHRRuqppzpJRmJhYRcYdBgwYxZcqUCp9/8OBBLBYLcXFxVVYmqVkUbqqaOhOLSD1lsVjK3caOHXter/vxxx/z17/+tcLnt2jRgsTERLp2rb4a9CFDhmCz2fjhhx+q7T3lNI2WqkqFeXBij7mvBTNFpJ5JTEx07i9btoy//OUv7Nmzx3nM39/f5fyCgoIKjZBp1KhRpcphs9mIjIys1DUXIj4+no0bNzJp0iQWLVrEJZdcUm3vXZqK3te6RDU3Ven4r+AoBL+GENLc06URkTrEMAyy8ws9shmGUaEyRkZGOreQkBAsFovzcW5uLg0bNuSDDz5g0KBB+Pn58e6775KSksLtt99O8+bNCQgIoFu3brz//vsur3t2s1SrVq34+9//zrhx4wgKCqJly5YsXLjQ+fzZzVJr1qzBYrHw7bff0qtXLwICAujfv79L8AKYNWsWTZo0ISgoiAkTJvDEE09w0UUXnfNzL1myhBEjRnD//fezbNkysrKyXJ5PTU3l3nvvJSIiAj8/P7p27cr//vc/5/Pr16/n8ssvJyAggNDQUIYOHcqpU6ecn3XevHkur3fRRRfxzDPPOB9bLBZef/11brjhBgIDA5k1axZ2u53x48cTExODv78/HTp04OWXXy5R9sWLF9OlSxd8fX2Jiopi0qRJAIwbN44RI0a4nFtYWEhkZCSLFy8+5z2pbqq5qUpnNklphIOIuFFOgZ3Of/nKI++9a+ZQAnzc8/Xx+OOPM3fuXJYsWYKvry+5ubnExsby+OOPExwczOeff87o0aNp3bo1ffv2LfN15s6dy1//+leefPJJPvroI+6//34GDhxIx44dy7xm+vTpzJ07l8aNGzNx4kTGjRvH+vXrAXjvvff429/+xvz587n00ktZunQpc+fOJSYmptzPYxgGS5Ys4Z///CcdO3akffv2fPDBB9x9992AOVHdsGHDyMjI4N1336VNmzbs2rULm80GQFxcHIMHD2bcuHG88soreHl5sXr16koPjX766aeZPXs2L730EjabDYfDQfPmzfnggw8IDw9nw4YN3HvvvURFRXHrrbcCsGDBAqZOncqcOXMYNmwYaWlpzvsxYcIEBg4cSGJiIlFRUQCsWLGCzMxM5/U1icJNVVJ/GxGRck2ZMoUbb7zR5dif/vQn5/7kyZP58ssv+fDDD8sNN8OHD+eBBx4AzMD00ksvsWbNmnLDzd/+9jcuv/xyAJ544gmuvfZacnNz8fPz49VXX2X8+PHOUPKXv/yFr7/+mszMzHI/zzfffEN2djZDhw4F4M4772TRokXO1/nmm2/46aef2L17N+3btwegdevWzuufe+45evXqxfz5853HunSp/Bxpf/zjHxk3bpzLsWeffda5HxMTw4YNG/jggw+c4WTWrFk8+uijPPzww87zevfuDUD//v3p0KED77zzDo899hhg1lDdcsstNGjQoNLlq2oKN1VJa0qJSBXx97axa+ZQj723u/Tq1cvlsd1uZ86cOSxbtoyjR4+Sl5dHXl4egYGB5b5O9+6nR6QWN38lJydX+Jri2ojk5GRatmzJnj17nGGpWJ8+fVi1alW5r7lo0SJGjRqFl5f59Xr77bfz5z//mT179tChQwfi4uJo3ry5M9icLS4ujltuuaXc96iIs+8rwOuvv84bb7zBoUOHyMnJIT8/39nMlpycTEJCAoMHDy7zNSdMmMDChQt57LHHSE5O5vPPP+fbb7+94LJWBYWbqmIYqrkRkSpjsVjc1jTkSWeHlrlz5/LSSy8xb948unXrRmBgIFOmTDnnSuhnd5i1WCznXGD0zGuKJ48785qzJ5Q7V1+jkydP8umnn1JQUMCCBQucx+12O4sXL+Yf//hHiU7UZzvX81artUQ5CgoKSpx39n394IMPeOSRR5g7dy79+vUjKCiI559/nh9//LFC7wtw11138cQTT7Bx40Y2btxIq1atGDBgwDmv8wR1KK4qqfGQlwZWbwjv4OnSiIjUCuvWreOGG27gzjvvpEePHrRu3Zq9e/dWezk6dOjATz/95HJs06ZN5V7z3nvv0bx5c7Zt20ZcXJxzmzdvHm+99RaFhYV0796dI0eO8Ntvv5X6Gt27dy+3NqRx48Yuo9DS09M5cODAOT/PunXr6N+/Pw888AAXX3wxbdu2Zd++fc7ng4KCaNWqVbnvHRYWxsiRI1myZAlLlixxNrXVRLU/9tdUx4qapBp3BC+tBSMiUhFt27Zl+fLlbNiwgdDQUF588UWSkpLo1KlTtZZj8uTJ3HPPPfTq1Yv+/fuzbNkytm/f7tI/5myLFi3i5ptvLjGfTnR0NI8//jiff/45N9xwAwMHDuSmm27ixRdfpG3btvz6669YLBauueYapk2bRrdu3XjggQeYOHEiPj4+rF69mltuuYXw8HCuvPJK3nzzTa677jpCQ0OZMWOGszNyedq2bcvbb7/NV199RUxMDO+88w4///yzSwfpZ555hokTJ9KkSRNnp+f169czefJk5zkTJkxgxIgR2O12xowZcx53tnp4vOZm/vz5xMTE4OfnR2xsLOvWrSvz3O+//55LL72UsLAw/P396dixIy+99FI1lrYS1CQlIlJpM2bMoGfPngwdOpRBgwYRGRnJyJEjq70cd9xxB9OmTeNPf/oTPXv25MCBA4wdOxY/P79Sz9+8eTPbtm3jpptuKvFcUFAQQ4YMYdGiRQAsX76c3r17c/vtt9O5c2cee+wx52io9u3b8/XXX7Nt2zb69OlDv379+L//+z9nH55p06YxcOBARowYwfDhwxk5ciRt2rQ55+eZOHEiN954I6NGjaJv376kpKSU6FM0ZswY5s2bx/z58+nSpQsjRowoUWt21VVXERUVxdChQ2natOm5b6SHWIyKTlhQBZYtW8bo0aOdQ+3+9a9/8cYbb7Br1y5atmxZ4vytW7fy66+/0r17dwIDA/n++++57777eOmll7j33nsr9J7p6emEhISQlpZGcHAVLmS59A749X8wdDb0e+Dc54uIlCM3N5cDBw44/xiU6nf11VcTGRnJO++84+mieEx2djZNmzZl8eLFJUa5uUN5v+eV+f72aLPUiy++yPjx45kwYQIA8+bN46uvvmLBggXMnj27xPkXX3wxF198sfNxq1at+Pjjj1m3bl2Fw0210YKZIiK1VnZ2Nq+//jpDhw7FZrPx/vvv880337By5UpPF80jHA4HSUlJzJ07l5CQEK6//npPF6lcHmuWys/PZ/PmzQwZMsTl+JAhQ9iwYUOFXmPr1q1s2LDBOU9BafLy8khPT3fZqlxOKqQeMve17IKISK1jsVhYsWIFAwYMIDY2lv/+978sX76cq666ytNF84j4+HiaNWvGBx98wOLFi53NZDWVx0p34sQJ7HY7ERERLscjIiJISkoq99rmzZtz/PhxCgsLeeaZZ5w1P6WZPXu2y8RF1eLYL+bPkBYQULk1UERExPP8/f355ptvPF2MGqNVq1YVXnajJvB4h+LS5hE4+9jZ1q1bx6ZNm3j99deZN29eiXVHzjRt2jTS0tKc2+HDh91S7nIVj5RSrY2IiEi181jNTXh4ODabrUQtTXJyconanLMVD13r1q0bx44d45lnnuH2228v9VxfX198fX3dU+iKStpu/tRIKRERkWrnsZobHx8fYmNjS3TOWrlyJf3796/w6xiGQV5enruLd2E0DFxERMRjPNojaOrUqYwePZpevXrRr18/Fi5cSHx8PBMnTgTMJqWjR4/y9ttvA/DPf/6Tli1bOhdC+/7773nhhRdcJhjyOHsBJP9q7ivciIiIVDuPhptRo0aRkpLCzJkzSUxMpGvXrqxYsYLo6GgAEhMTiY+Pd57vcDiYNm0aBw4cwMvLizZt2jBnzhzuu+8+T32Ekk7sBXse+ARBw2hPl0ZERKTe8egkfp5Q5ZP4bVsGn9wLLfvBuC/d//oiUi9pEj+pD9w1iZ/HR0vVOcfU30ZExJ0GDRrElClTnI9btWrFvHnzyr3GYrHw6aefXvB7u+t1pHop3LhbcWdiDQMXkXruuuuuK3PSu40bN2KxWNiyZUulX/fnn392+6z0zzzzDBdddFGJ44mJiQwbNsyt71WWnJwcQkNDadSoETk5OdXynnWVwo07GYZGSomIFBk/fjyrVq3i0KFDJZ5bvHgxF110ET179qz06zZu3JiAgAB3FPGcIiMjq206keXLl9O1a1c6d+7Mxx9/XC3vWRbDMCgsLPRoGS6Ewo07ZSRCdgpYbNCkk6dLIyJ1mWFAfpZntgp21RwxYgRNmjThzTffdDmenZ3NsmXLGD9+PCkpKdx+++00b96cgIAAunXrVu7ErFCyWWrv3r0MHDgQPz8/OnfuXOr6T48//jjt27cnICCA1q1bM2PGDAoKCgB48803efbZZ9m2bRsWiwWLxeIs89nNUjt27ODKK6/E39+fsLAw7r33XjIzM53Pjx07lpEjR/LCCy8QFRVFWFgYDz74oPO9yrNo0SLuvPNO7rzzTucK4mf65ZdfuPbaawkODiYoKIgBAwawb98+5/OLFy+mS5cu+Pr6EhUVxaRJkwA4ePAgFouFuLg457mpqalYLBbWrFkDwJo1a7BYLHz11Vf06tULX19f1q1bx759+7jhhhuIiIigQYMG9O7du8TMzXl5eTz22GO0aNECX19f2rVrx6JFizAMg7Zt2/LCCy+4nL9z506sVqtL2d2tZi8OUdskFc1MHN4OvP09WxYRqdsKsuHvTT3z3k8mgE/gOU/z8vLirrvu4s033+Qvf/mLc/b5Dz/8kPz8fO644w6ys7OJjY3l8ccfJzg4mM8//5zRo0fTunVr+vbte873cDgc3HjjjYSHh/PDDz+Qnp7u0j+nWFBQEG+++SZNmzZlx44d3HPPPQQFBfHYY48xatQodu7cyZdffun84g4JCSnxGtnZ2VxzzTVccskl/PzzzyQnJzNhwgQmTZrkEuBWr15NVFQUq1ev5vfff2fUqFFcdNFF3HPPPWV+jn379rFx40Y+/vhjDMNgypQp7N+/n9atWwNw9OhRBg4cyKBBg1i1ahXBwcGsX7/eWbuyYMECpk6dypw5cxg2bBhpaWmsX7/+nPfvbI899hgvvPACrVu3pmHDhhw5coThw4cza9Ys/Pz8eOutt7juuuvYs2cPLVu2BOCuu+5i48aNvPLKK/To0YMDBw5w4sQJLBYL48aNY8mSJfzpT39yvsfixYsZMGAAbdq0qXT5Kkrhxp00M7GIiItx48bx/PPPs2bNGq644grA/HK78cYbCQ0NJTQ01OWLb/LkyXz55Zd8+OGHFQo333zzDbt37+bgwYM0b94cgL///e8l+sk89dRTzv1WrVrx6KOPsmzZMh577DH8/f1p0KABXl5eREZGlvle7733Hjk5Obz99tsEBprh7rXXXuO6667jH//4h3N2/dDQUF577TVsNhsdO3bk2muv5dtvvy033CxevJhhw4YRGhoKwDXXXMPixYuZNWsWYM7zFhISwtKlS/H29gagffv2zutnzZrFo48+ysMPP+w81rt373Pev7PNnDmTq6++2vk4LCyMHj16uLzPJ598wmeffcakSZP47bff+OCDD1i5cqWzf1VxIAO4++67+ctf/sJPP/1Enz59KCgo4N133+X555+vdNkqQ+HGndTfRkSqi3eAWYPiqfeuoI4dO9K/f38WL17MFVdcwb59+1i3bh1ff/01AHa7nTlz5rBs2TKOHj1KXl4eeXl5zvBwLrt376Zly5bOYAPQr1+/Eud99NFHzJs3j99//53MzEwKCwsrPR3I7t276dGjh0vZLr30UhwOB3v27HGGmy5dumCz2ZznREVFsWPHjjJf126389Zbb/Hyyy87j91555088sgjPPvss9hsNuLi4hgwYIAz2JwpOTmZhIQEBg8eXKnPU5pevXq5PM7KyuLZZ5/lf//7HwkJCRQWFpKTk+Ocgy4uLg6bzcbll19e6utFRUVx7bXXsnjxYvr06cP//vc/cnNzueWWWy64rOVRnxt3Kl4wU+FGRKqaxWI2DXliO8fixmcbP348y5cvJz09nSVLlhAdHe38Ip47dy4vvfQSjz32GKtWrSIuLo6hQ4eSn59fodcubaq2sxdf/uGHH7jtttsYNmwY//vf/9i6dSvTp0+v8Huc+V5lLex85vGzA4jFYsHhcJT5ul999RVHjx5l1KhReHl54eXlxW233caRI0ecIdDfv+yuDuU9B2C1Wp3lL1ZWH6CzQ+Wf//xnli9fzt/+9jfWrVtHXFwc3bp1c967c703wIQJE1i6dCk5OTksWbKEUaNGVXmHcIUbd8nLhJSizlERCjciIsVuvfVWbDYb//nPf3jrrbe4++67nWFg3bp13HDDDdx555306NGD1q1bs3fv3gq/dufOnYmPjych4XQt1saNG13OWb9+PdHR0UyfPp1evXrRrl27EiO4fHx8sNvt53yvuLg4srKyXF7barW6NBFV1qJFi7jtttuIi4tz2e644w5nx+Lu3buzbt26UkNJUFAQrVq14ttvvy319Rs3bgyYw9qLndm5uDzr1q1j7Nix/OEPf6Bbt25ERkZy8OBB5/PdunXD4XCwdu3aMl9j+PDhBAYGsmDBAr744gvGjRtXofe+EAo37pKeAEGR0CASGjT2dGlERGqMBg0aMGrUKJ588kkSEhIYO3as87m2bduycuVKNmzYwO7du7nvvvtISkqq8GtfddVVdOjQgbvuuott27axbt06pk+f7nJO27ZtiY+PZ+nSpezbt49XXnmFTz75xOWcVq1aceDAAeLi4jhx4kSpCzLfcccd+Pn5MWbMGHbu3Mnq1auZPHkyo0ePdjZJVdbx48f573//y5gxY+jatavLNmbMGD777DOOHz/OpEmTSE9P57bbbmPTpk3s3buXd955hz179gDmPD1z587llVdeYe/evWzZsoVXX30VMGtXLrnkEubMmcOuXbv47rvvXPogladt27Z8/PHHxMXFsW3bNv74xz+61EK1atWKMWPGMG7cOD799FMOHDjAmjVr+OCDD5zn2Gw2xo4dy7Rp02jbtm2pzYbupnDjLo3bw6O/wqSfPV0SEZEaZ/z48Zw6dYqrrrrKOcoGYMaMGfTs2ZOhQ4cyaNAgIiMjGTlyZIVf12q18sknn5CXl0efPn2YMGECf/vb31zOueGGG3jkkUeYNGkSF110ERs2bGDGjBku59x0001cc801XHHFFTRu3LjU4egBAQF89dVXnDx5kt69e3PzzTczePBgXnvttcrdjDMUd04urb/MFVdcQVBQEO+88w5hYWGsWrWKzMxMLr/8cmJjY/n3v//tbAIbM2YM8+bNY/78+XTp0oURI0a41IAtXryYgoICevXqxcMPP+zsqHwuL730EqGhofTv35/rrruOoUOHlpibaMGCBdx888088MADdOzYkXvuuceldgvM//75+fnVUmsDWlvK08UREakQrS0ltdn69esZNGgQR44cKbeWy11rS2m0lIiIiFSJvLw8Dh8+zIwZM7j11lvPu/mustQsJSIiIlXi/fffp0OHDqSlpfHcc89V2/sq3IiIiEiVGDt2LHa7nc2bN9OsWbNqe1+FGxEREalTFG5ERGqRejYGROoZd/1+K9yIiNQCxdP5V3ZWXZHapPj3+8zlK86HRkuJiNQCXl5eBAQEcPz4cby9vZ1T6ovUFQ6Hg+PHjxMQEICX14XFE4UbEZFawGKxEBUVxYEDB0osHSBSV1itVlq2bFnmGl4VpXAjIlJL+Pj40K5dOzVNSZ3l4+PjllpJhRsRkVrEarVqhmKRc1CjrYiIiNQpCjciIiJSpyjciIiISJ1S7/rcFE8QlJ6e7uGSiIiISEUVf29XZKK/ehduMjIyAGjRooWHSyIiIiKVlZGRQUhISLnnWIx6Npe3w+EgISGBoKCgCx5Hf7b09HRatGjB4cOHCQ4OdutrS0m639VL97t66X5XL93v6nU+99swDDIyMmjatOk5h4vXu5obq9VK8+bNq/Q9goOD9T9HNdL9rl6639VL97t66X5Xr8re73PV2BRTh2IRERGpUxRuREREpE5RuHEjX19fnn76aXx9fT1dlHpB97t66X5XL93v6qX7Xb2q+n7Xuw7FIiIiUrep5kZERETqFIUbERERqVMUbkRERKROUbgRERGROkXhxk3mz59PTEwMfn5+xMbGsm7dOk8Xqc747rvvuO6662jatCkWi4VPP/3U5XnDMHjmmWdo2rQp/v7+DBo0iF9++cUzha3lZs+eTe/evQkKCqJJkyaMHDmSPXv2uJyj++0+CxYsoHv37s6JzPr168cXX3zhfF73umrNnj0bi8XClClTnMd0z93nmWeewWKxuGyRkZHO56vyXivcuMGyZcuYMmUK06dPZ+vWrQwYMIBhw4YRHx/v6aLVCVlZWfTo0YPXXnut1Oefe+45XnzxRV577TV+/vlnIiMjufrqq53riEnFrV27lgcffJAffviBlStXUlhYyJAhQ8jKynKeo/vtPs2bN2fOnDls2rSJTZs2ceWVV3LDDTc4/4HXva46P//8MwsXLqR79+4ux3XP3atLly4kJiY6tx07djifq9J7bcgF69OnjzFx4kSXYx07djSeeOIJD5Wo7gKMTz75xPnY4XAYkZGRxpw5c5zHcnNzjZCQEOP111/3QAnrluTkZAMw1q5daxiG7nd1CA0NNd544w3d6yqUkZFhtGvXzli5cqVx+eWXGw8//LBhGPr9drenn37a6NGjR6nPVfW9Vs3NBcrPz2fz5s0MGTLE5fiQIUPYsGGDh0pVfxw4cICkpCSX++/r68vll1+u++8GaWlpADRq1AjQ/a5KdrudpUuXkpWVRb9+/XSvq9CDDz7Itddey1VXXeVyXPfc/fbu3UvTpk2JiYnhtttuY//+/UDV3+t6t3Cmu504cQK73U5ERITL8YiICJKSkjxUqvqj+B6Xdv8PHTrkiSLVGYZhMHXqVC677DK6du0K6H5XhR07dtCvXz9yc3Np0KABn3zyCZ07d3b+A6977V5Lly5ly5Yt/PzzzyWe0++3e/Xt25e3336b9u3bc+zYMWbNmkX//v355ZdfqvxeK9y4icVicXlsGEaJY1J1dP/db9KkSWzfvp3vv/++xHO63+7ToUMH4uLiSE1NZfny5YwZM4a1a9c6n9e9dp/Dhw/z8MMP8/XXX+Pn51fmebrn7jFs2DDnfrdu3ejXrx9t2rThrbfe4pJLLgGq7l6rWeoChYeHY7PZStTSJCcnl0ik4n7FPe91/91r8uTJfPbZZ6xevZrmzZs7j+t+u5+Pjw9t27alV69ezJ49mx49evDyyy/rXleBzZs3k5ycTGxsLF5eXnh5ebF27VpeeeUVvLy8nPdV97xqBAYG0q1bN/bu3Vvlv98KNxfIx8eH2NhYVq5c6XJ85cqV9O/f30Olqj9iYmKIjIx0uf/5+fmsXbtW9/88GIbBpEmT+Pjjj1m1ahUxMTEuz+t+Vz3DMMjLy9O9rgKDBw9mx44dxMXFObdevXpxxx13EBcXR+vWrXXPq1BeXh67d+8mKiqq6n+/L7hLshhLly41vL29jUWLFhm7du0ypkyZYgQGBhoHDx70dNHqhIyMDGPr1q3G1q1bDcB48cUXja1btxqHDh0yDMMw5syZY4SEhBgff/yxsWPHDuP22283oqKijPT0dA+XvPa5//77jZCQEGPNmjVGYmKic8vOznaeo/vtPtOmTTO+++4748CBA8b27duNJ5980rBarcbXX39tGIbudXU4c7SUYeieu9Ojjz5qrFmzxti/f7/xww8/GCNGjDCCgoKc341Vea8Vbtzkn//8pxEdHW34+PgYPXv2dA6dlQu3evVqAyixjRkzxjAMc0jh008/bURGRhq+vr7GwIEDjR07dni20LVUafcZMJYsWeI8R/fbfcaNG+f8d6Nx48bG4MGDncHGMHSvq8PZ4Ub33H1GjRplREVFGd7e3kbTpk2NG2+80fjll1+cz1flvbYYhmFceP2PiIiISM2gPjciIiJSpyjciIiISJ2icCMiIiJ1isKNiIiI1CkKNyIiIlKnKNyIiIhInaJwIyIiInWKwo2IiIjUKQo3IlKjWCwWPv30U08Xo1LWrFmDxWIhNTXV00URERRuRKTI2LFjsVgsJbZrrrnG00U7p0GDBmGxWFi6dKnL8Xnz5tGqVSvPFEpEPEbhRkScrrnmGhITE122999/39PFqhA/Pz+eeuopCgoKPF0Ut8nPz/d0EURqJYUbEXHy9fUlMjLSZQsNDXU+b7FYWLBgAcOGDcPf35+YmBg+/PBDl9fYsWMHV155Jf7+/oSFhXHvvfeSmZnpcs7ixYvp0qULvr6+REVFMWnSJJfnT5w4wR/+8AcCAgJo164dn3322TnLfvvtt5OWlsa///3vMs8ZO3YsI0eOdDk2ZcoUBg0a5Hw8aNAgJk+ezJQpUwgNDSUiIoKFCxeSlZXF3XffTVBQEG3atOGLL74o8frr16+nR48e+Pn50bdvX3bs2OHy/IYNGxg4cCD+/v60aNGChx56iKysLOfzrVq1YtasWYwdO5aQkBDuueeec35uESlJ4UZEKmXGjBncdNNNbNu2jTvvvJPbb7+d3bt3A5Cdnc0111xDaGgoP//8Mx9++CHffPONS3hZsGABDz74IPfeey87duzgs88+o23bti7v8eyzz3Lrrbeyfft2hg8fzh133MHJkyfLLVdwcDBPPvkkM2fOdAkM5+Ott94iPDycn376icmTJ3P//fdzyy230L9/f7Zs2cLQoUMZPXo02dnZLtf9+c9/5oUXXuDnn3+mSZMmXH/99c6apB07djB06FBuvPFGtm/fzrJly/j+++9LBLvnn3+erl27snnzZmbMmHFBn0Ok3nLL2uIiUuuNGTPGsNlsRmBgoMs2c+ZM5zmAMXHiRJfr+vbta9x///2GYRjGwoULjdDQUCMzM9P5/Oeff25YrVYjKSnJMAzDaNq0qTF9+vQyywEYTz31lPNxZmamYbFYjC+++KLMay6//HLj4YcfNnJzc43o6GhnmV966SUjOjra5TPecMMNLtc+/PDDxuWXX+7yWpdddpnzcWFhoREYGGiMHj3aeSwxMdEAjI0bNxqGYRirV682AGPp0qXOc1JSUgx/f39j2bJlhmEYxujRo417773X5b3XrVtnWK1WIycnxzAMw4iOjjZGjhxZ5ucUkYrx8my0EpGa5IorrmDBggUuxxo1auTyuF+/fiUex8XFAbB792569OhBYGCg8/lLL70Uh8PBnj17sFgsJCQkMHjw4HLL0b17d+d+YGAgQUFBJCcnn7P8vr6+zJw5k0mTJnH//fef8/yKvL/NZiMsLIxu3bo5j0VERACUKNOZ96ZRo0Z06NDBWau1efNmfv/9d9577z3nOYZh4HA4OHDgAJ06dQKgV69e511uETEp3IiIU2BgYIkmooqwWCyA+WVdvF/aOf7+/hV6PW9v7xLXOhyOCl1755138sILLzBr1qwSI6WsViuGYbgcK60Dcmnvf+ax4s9YkTKdee59993HQw89VOKcli1bOvfPDIYicn7U50ZEKuWHH34o8bhjx44AdO7cmbi4OJc+L+vXr8dqtdK+fXuCgoJo1aoV3377bZWVz2q1Mnv2bBYsWMDBgwddnmvcuDGJiYkux4prndzhzHtz6tQpfvvtN+e96dmzJ7/88gtt27Ytsfn4+LitDCKicCMiZ8jLyyMpKcllO3HihMs5H374IYsXL+a3337j6aef5qeffnJ2ir3jjjvw8/NjzJgx7Ny5k9WrVzN58mRGjx7tbMp55plnmDt3Lq+88gp79+5ly5YtvPrqq279HNdeey19+/blX//6l8vxK6+8kk2bNvH222+zd+9enn76aXbu3Om29505cybffvstO3fuZOzYsYSHhztHZz3++ONs3LiRBx98kLi4OPbu3ctnn33G5MmT3fb+ImJSuBERpy+//JKoqCiX7bLLLnM559lnn2Xp0qV0796dt956i/fee4/OnTsDEBAQwFdffcXJkyfp3bs3N998M4MHD+a1115zXj9mzBjmzZvH/Pnz6dKlCyNGjGDv3r1u/yz/+Mc/yM3NdTk2dOhQZsyYwWOPPUbv3r3JyMjgrrvuctt7zpkzh4cffpjY2FgSExP57LPPnLUy3bt3Z+3atezdu5cBAwZw8cUXM2PGDKKiotz2/iJishhnN0CLiJTBYrHwySeflJgrRkSkJlHNjYiIiNQpCjciIiJSp2gouIhUmFqxRaQ2UM2NiIiI1CkKNyIiIlKnKNyIiIhInaJwIyIiInWKwo2IiIjUKQo3IiIiUqco3IiIiEidonAjIiIidcr/A6eqn3/uEtk2AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = [i for i in range(no_epochs)]\n",
    "# plot lines\n",
    "plt.plot(x, train_accs, label = \"Training Accuracy\")\n",
    "plt.plot(x, val_accs, label = \"Validation Accuracy\")\n",
    "plt.xlabel(\"Epoch Number\")\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "d2f44392",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:45:59.198916Z",
     "iopub.status.busy": "2023-03-25T16:45:59.198552Z",
     "iopub.status.idle": "2023-03-25T16:45:59.402272Z",
     "shell.execute_reply": "2023-03-25T16:45:59.401281Z"
    },
    "papermill": {
     "duration": 1.151584,
     "end_time": "2023-03-25T16:45:59.404377",
     "exception": false,
     "start_time": "2023-03-25T16:45:58.252793",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = [i for i in range(no_epochs)]\n",
    "# plot lines\n",
    "plt.plot(x, train_losses, label = \"Training Loss\")\n",
    "plt.plot(x, val_losses, label = \"Validation Loss\")\n",
    "plt.xlabel(\"Epoch Number\")\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "66c26624",
   "metadata": {
    "papermill": {
     "duration": 0.981953,
     "end_time": "2023-03-25T16:46:01.275030",
     "exception": false,
     "start_time": "2023-03-25T16:46:00.293077",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Test and Dev Accuracies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "04b9f9a5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:46:03.313119Z",
     "iopub.status.busy": "2023-03-25T16:46:03.312761Z",
     "iopub.status.idle": "2023-03-25T16:46:59.584895Z",
     "shell.execute_reply": "2023-03-25T16:46:59.583565Z"
    },
    "papermill": {
     "duration": 57.263567,
     "end_time": "2023-03-25T16:46:59.587031",
     "exception": false,
     "start_time": "2023-03-25T16:46:02.323464",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 283/283 [00:56<00:00,  5.03it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dev Loss: 0.0001\n",
      "Dev Accuracy of  fake: 95% (6641/6930)\n",
      "Dev Accuracy of  true: 89% (1870/2094)\n",
      "Dev Accuracy of 94% (8511/9024)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "dev_loss = 0.0\n",
    "class_correct = list(0 for i in range(len(classes)))\n",
    "class_total = list(0 for i in range(len(classes)))\n",
    "model_ft.eval()\n",
    "dev_predictions = []\n",
    "dev_targets = []\n",
    "\n",
    "for data, target in tqdm(dev_loader):\n",
    "    data, target = data.to(device), target.to(device)\n",
    "    with torch.no_grad(): # turn off autograd for faster testing\n",
    "        output = model_ft(data)\n",
    "        loss = criterion(output, target)\n",
    "    dev_loss = loss.item() * data.size(0)\n",
    "    _, pred = torch.max(output, 1)\n",
    "    pred_list = np.squeeze(pred.cpu().numpy())\n",
    "    target_list = np.squeeze(target.cpu().numpy())\n",
    "    dev_predictions.extend(pred_list)\n",
    "    dev_targets.extend(target_list)\n",
    "    correct_tensor = pred.eq(target.data.view_as(pred))\n",
    "    correct = np.squeeze(correct_tensor.cpu().numpy())\n",
    "#     print('target',target)\n",
    "#     print('##################################################')\n",
    "#     print('output',output)\n",
    "#     print('##################################################')\n",
    "#     print('pred',pred)\n",
    "#     print('##################################################')\n",
    "#     print('correct_tensor',correct_tensor)\n",
    "#     print('##################################################')\n",
    "#     print('correct',correct)\n",
    "    if len(target) == 32:\n",
    "        for i in range(32):\n",
    "            label = target.data[i]\n",
    "            class_correct[label] += correct[i].item()\n",
    "            class_total[label] += 1\n",
    "\n",
    "# print('class_correct',class_correct)\n",
    "# print('##################################################')\n",
    "# print('class_total',class_total)\n",
    "\n",
    "dev_loss = dev_loss / dev_data_len\n",
    "print('Dev Loss: {:.4f}'.format(dev_loss))\n",
    "for i in range(len(classes)):\n",
    "    if class_total[i] > 0:\n",
    "        print(\"Dev Accuracy of %5s: %2d%% (%2d/%2d)\" % (\n",
    "            classes[i], 100*class_correct[i]/class_total[i], np.sum(class_correct[i]), np.sum(class_total[i])\n",
    "        ))\n",
    "    else:\n",
    "        print(\"Dev accuracy of %5s: NA\" % (classes[i]))\n",
    "print(\"Dev Accuracy of %2d%% (%2d/%2d)\" % (\n",
    "            100*np.sum(class_correct)/np.sum(class_total), np.sum(class_correct), np.sum(class_total)\n",
    "        ))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "1d50f3fb",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:47:01.444923Z",
     "iopub.status.busy": "2023-03-25T16:47:01.444477Z",
     "iopub.status.idle": "2023-03-25T16:47:49.871278Z",
     "shell.execute_reply": "2023-03-25T16:47:49.870176Z"
    },
    "papermill": {
     "duration": 49.382863,
     "end_time": "2023-03-25T16:47:49.873755",
     "exception": false,
     "start_time": "2023-03-25T16:47:00.490892",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 243/243 [00:48<00:00,  5.02it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Loss: 0.0011\n",
      "Test Accuracy of  fake: 84% (5193/6168)\n",
      "Test Accuracy of  true: 84% (1336/1576)\n",
      "Test Accuracy of 84% (6529/7744)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "test_loss = 0.0\n",
    "class_correct = list(0 for i in range(len(classes)))\n",
    "class_total = list(0 for i in range(len(classes)))\n",
    "model_ft.eval()\n",
    "test_predictions = []\n",
    "test_targets = []\n",
    "\n",
    "for data, target in tqdm(test_loader):\n",
    "    data, target = data.to(device), target.to(device)\n",
    "    with torch.no_grad(): # turn off autograd for faster testing\n",
    "        output = model_ft(data)\n",
    "        loss = criterion(output, target)\n",
    "    test_loss = loss.item() * data.size(0)\n",
    "    _, pred = torch.max(output, 1)\n",
    "    pred_list = np.squeeze(pred.cpu().numpy())\n",
    "    target_list = np.squeeze(target.cpu().numpy())\n",
    "    test_predictions.extend(pred_list)\n",
    "    test_targets.extend(target_list)\n",
    "    correct_tensor = pred.eq(target.data.view_as(pred))\n",
    "    correct = np.squeeze(correct_tensor.cpu().numpy())\n",
    "#     print('target',target)\n",
    "#     print('##################################################')\n",
    "#     print('output',output)\n",
    "#     print('##################################################')\n",
    "#     print('pred',pred)\n",
    "#     print('##################################################')\n",
    "#     print('correct_tensor',correct_tensor)\n",
    "#     print('##################################################')\n",
    "#     print('correct',correct)\n",
    "    if len(target) == 32:\n",
    "        for i in range(32):\n",
    "            label = target.data[i]\n",
    "            class_correct[label] += correct[i].item()\n",
    "            class_total[label] += 1\n",
    "\n",
    "# print('class_correct',class_correct)\n",
    "# print('##################################################')\n",
    "# print('class_total',class_total)\n",
    "\n",
    "test_loss = test_loss / test_data_len\n",
    "print('Test Loss: {:.4f}'.format(test_loss))\n",
    "for i in range(len(classes)):\n",
    "    if class_total[i] > 0:\n",
    "        print(\"Test Accuracy of %5s: %2d%% (%2d/%2d)\" % (\n",
    "            classes[i], 100*class_correct[i]/class_total[i], np.sum(class_correct[i]), np.sum(class_total[i])\n",
    "        ))\n",
    "    else:\n",
    "        print(\"Test accuracy of %5s: NA\" % (classes[i]))\n",
    "print(\"Test Accuracy of %2d%% (%2d/%2d)\" % (\n",
    "            100*np.sum(class_correct)/np.sum(class_total), np.sum(class_correct), np.sum(class_total)\n",
    "        ))\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e7a6b2da",
   "metadata": {
    "papermill": {
     "duration": 1.174697,
     "end_time": "2023-03-25T16:47:51.993369",
     "exception": false,
     "start_time": "2023-03-25T16:47:50.818672",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Save the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "c6f36f7f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:47:53.896007Z",
     "iopub.status.busy": "2023-03-25T16:47:53.895623Z",
     "iopub.status.idle": "2023-03-25T16:47:57.460787Z",
     "shell.execute_reply": "2023-03-25T16:47:57.459563Z"
    },
    "papermill": {
     "duration": 4.556607,
     "end_time": "2023-03-25T16:47:57.463411",
     "exception": false,
     "start_time": "2023-03-25T16:47:52.906804",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "example = torch.rand(1, 3, 224, 224)\n",
    "traced_script_module = torch.jit.trace(model.cpu(), example)\n",
    "traced_script_module.save(\"vit_model.pt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "a89aeef6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:47:59.365023Z",
     "iopub.status.busy": "2023-03-25T16:47:59.364658Z",
     "iopub.status.idle": "2023-03-25T16:47:59.368950Z",
     "shell.execute_reply": "2023-03-25T16:47:59.367943Z"
    },
    "papermill": {
     "duration": 0.901826,
     "end_time": "2023-03-25T16:47:59.371035",
     "exception": false,
     "start_time": "2023-03-25T16:47:58.469209",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# print('Target Labels: ',len(targets))\n",
    "# print('Prediction Labels: ',len(predictions))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aa571f14",
   "metadata": {
    "papermill": {
     "duration": 0.981545,
     "end_time": "2023-03-25T16:48:01.326975",
     "exception": false,
     "start_time": "2023-03-25T16:48:00.345430",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# Other Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "c3519c0d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:48:03.943219Z",
     "iopub.status.busy": "2023-03-25T16:48:03.942275Z",
     "iopub.status.idle": "2023-03-25T16:48:03.958593Z",
     "shell.execute_reply": "2023-03-25T16:48:03.957332Z"
    },
    "papermill": {
     "duration": 1.036706,
     "end_time": "2023-03-25T16:48:03.960740",
     "exception": false,
     "start_time": "2023-03-25T16:48:02.924034",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The EER for Dev is: 7.44\n",
      "The EER for Test is: 15.54\n"
     ]
    }
   ],
   "source": [
    "#EER\n",
    "def compute_eer(label, pred, positive_label=1):\n",
    "    # all fpr, tpr, fnr, fnr, threshold are lists (in the format of np.array)\n",
    "    fpr, tpr, threshold = sklearn.metrics.roc_curve(label, pred)\n",
    "    fnr = 1 - tpr\n",
    "\n",
    "    # the threshold of fnr == fpr\n",
    "    eer_threshold = threshold[np.nanargmin(np.absolute((fnr - fpr)))]\n",
    "\n",
    "    # theoretically eer from fpr and eer from fnr should be identical but they can be slightly differ in reality\n",
    "    eer_1 = fpr[np.nanargmin(np.absolute((fnr - fpr)))]\n",
    "    eer_2 = fnr[np.nanargmin(np.absolute((fnr - fpr)))]\n",
    "\n",
    "    # return the mean of eer from fpr and from fnr\n",
    "    eer = (eer_1 + eer_2) / 2\n",
    "   \n",
    "    return eer\n",
    "\n",
    "test_EER = compute_eer(test_targets,test_predictions)\n",
    "dev_EER = compute_eer(dev_targets,dev_predictions)\n",
    "\n",
    "print('The EER for Dev is: {:.2f}'.format(dev_EER*100))\n",
    "print('The EER for Test is: {:.2f}'.format(test_EER*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "0e9c7728",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:48:05.818864Z",
     "iopub.status.busy": "2023-03-25T16:48:05.818507Z",
     "iopub.status.idle": "2023-03-25T16:48:05.837831Z",
     "shell.execute_reply": "2023-03-25T16:48:05.836544Z"
    },
    "papermill": {
     "duration": 0.978669,
     "end_time": "2023-03-25T16:48:05.840138",
     "exception": false,
     "start_time": "2023-03-25T16:48:04.861469",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The HTER for Dev is: 7.44\n",
      "The HTER for Test is: 15.54\n"
     ]
    }
   ],
   "source": [
    "#HTER\n",
    "def calculate_HTER(y_true, y_pred):\n",
    "    cm = confusion_matrix(y_true, y_pred)\n",
    "    tn, fp, fn, tp = cm.ravel()\n",
    "    far = fp / (fp + tn)\n",
    "    frr = fn / (tp + fn)\n",
    "    hter = (far + frr) / 2\n",
    "    return hter\n",
    "\n",
    "dev_hter = calculate_HTER(dev_targets, dev_predictions)\n",
    "print('The HTER for Dev is: {:.2f}'.format(dev_hter*100))\n",
    "\n",
    "test_hter = calculate_HTER(test_targets, test_predictions)\n",
    "print('The HTER for Test is: {:.2f}'.format(test_hter*100))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "a1e402e5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:48:07.750340Z",
     "iopub.status.busy": "2023-03-25T16:48:07.749970Z",
     "iopub.status.idle": "2023-03-25T16:48:07.793406Z",
     "shell.execute_reply": "2023-03-25T16:48:07.792172Z"
    },
    "papermill": {
     "duration": 1.003856,
     "end_time": "2023-03-25T16:48:07.796005",
     "exception": false,
     "start_time": "2023-03-25T16:48:06.792149",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "######## Dev Oulu Metrics ##########\n",
      "APCER: 4.18\n",
      "BPCER: 10.69\n",
      "ACER: 7.44\n",
      "######## Test Oulu Metrics ##########\n",
      "APCER: 15.81\n",
      "BPCER: 15.26\n",
      "ACER: 15.54\n"
     ]
    }
   ],
   "source": [
    "#APCER, BPCER, ACER\n",
    "test_attack_types = [ (1 if x==1 else 2) for x in test_targets]\n",
    "dev_attack_types = [ (1 if x==1 else 2) for x in dev_targets]\n",
    "\n",
    "\n",
    "# returns the metrics APCER, BPCER and ACER\n",
    "test_apcer, test_bpcer, test_acer = oulumetrics.calculate_metrics(test_attack_types, test_predictions)\n",
    "dev_apcer, dev_bpcer, dev_acer = oulumetrics.calculate_metrics(dev_attack_types, dev_predictions)\n",
    "\n",
    "print('######## Dev Oulu Metrics ##########')\n",
    "print('APCER: {:.2f}'.format(dev_apcer*100))\n",
    "print('BPCER: {:.2f}'.format(dev_bpcer*100))\n",
    "print('ACER: {:.2f}'.format(dev_acer*100))\n",
    "\n",
    "print('######## Test Oulu Metrics ##########')\n",
    "print('APCER: {:.2f}'.format(test_apcer*100))\n",
    "print('BPCER: {:.2f}'.format(test_bpcer*100))\n",
    "print('ACER: {:.2f}'.format(test_acer*100))\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "a0eba3f7",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:48:09.660541Z",
     "iopub.status.busy": "2023-03-25T16:48:09.660163Z",
     "iopub.status.idle": "2023-03-25T16:48:09.680372Z",
     "shell.execute_reply": "2023-03-25T16:48:09.679351Z"
    },
    "papermill": {
     "duration": 0.979866,
     "end_time": "2023-03-25T16:48:09.682536",
     "exception": false,
     "start_time": "2023-03-25T16:48:08.702670",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix:\n",
      "[[5202  977]\n",
      " [ 241 1338]]\n",
      "F1 Score: 0.6872110939907549\n"
     ]
    }
   ],
   "source": [
    "# Compute confusion matrix\n",
    "cm = confusion_matrix(test_targets, test_predictions)\n",
    "print(\"Confusion Matrix:\")\n",
    "print(cm)\n",
    "\n",
    "# Compute F1 score\n",
    "f1 = f1_score(test_targets, test_predictions)\n",
    "print(\"F1 Score:\", f1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "c86772c0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-03-25T16:48:11.577502Z",
     "iopub.status.busy": "2023-03-25T16:48:11.576609Z",
     "iopub.status.idle": "2023-03-25T16:48:11.820221Z",
     "shell.execute_reply": "2023-03-25T16:48:11.819258Z"
    },
    "papermill": {
     "duration": 1.208883,
     "end_time": "2023-03-25T16:48:11.822246",
     "exception": false,
     "start_time": "2023-03-25T16:48:10.613363",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f585c08a6d0>"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Compute ROC curve\n",
    "fpr, tpr, thresholds = roc_curve(test_targets, test_predictions)\n",
    "\n",
    "# Plot ROC curve\n",
    "plt.plot(fpr, tpr, color='blue', label='ROC Curve')\n",
    "plt.plot([0, 1], [0, 1], color='red', linestyle='--', label='Random Guess')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('Receiver Operating Characteristic (ROC) Curve')\n",
    "plt.legend()\n",
    "\n",
    "# # Add predicted labels to plot\n",
    "# for i, threshold in enumerate(thresholds):\n",
    "#     plt.annotate(threshold, (fpr[i], tpr[i]), textcoords=\"offset points\", xytext=(0,10), ha='center')\n",
    "\n",
    "# for i, label in enumerate(test_predictions):\n",
    "#     if label == 1:\n",
    "#         plt.plot(fpr[i], tpr[i], 'o', color='green')\n",
    "#     else:\n",
    "#         plt.plot(fpr[i], tpr[i], 'o', color='red')\n",
    "\n",
    "# plt.show()\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 4732.643157,
   "end_time": "2023-03-25T16:48:14.748119",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2023-03-25T15:29:22.104962",
   "version": "2.4.0"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {
     "0354f582f33b449a82b70a7aa9766fa1": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_41a3543511e3410db965bbbd2aadcf45",
        "IPY_MODEL_bffa8c42220a42f6804642d9265b0a2a",
        "IPY_MODEL_1bd18d286fe346daa0ec6bb0460fc4f0"
       ],
       "layout": "IPY_MODEL_522432dc94bd4e57835f2d59cbcb5afa"
      }
     },
     "1bd18d286fe346daa0ec6bb0460fc4f0": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_29879462e93c40468cbdac6bce8deb40",
       "placeholder": "​",
       "style": "IPY_MODEL_b9c99fae866d44889a3672917fec0e21",
       "value": " 330M/330M [00:02&lt;00:00, 127MB/s]"
      }
     },
     "29879462e93c40468cbdac6bce8deb40": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "3378c8b694544338b1766b1a252866b0": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "41a3543511e3410db965bbbd2aadcf45": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_3378c8b694544338b1766b1a252866b0",
       "placeholder": "​",
       "style": "IPY_MODEL_a6089a2e0ccd449b81cf1adee2ef2eb3",
       "value": "100%"
      }
     },
     "522432dc94bd4e57835f2d59cbcb5afa": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "a6089a2e0ccd449b81cf1adee2ef2eb3": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "DescriptionStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "DescriptionStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "description_width": ""
      }
     },
     "acb8645a19cf419d9d4d2cfdac027622": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": ""
      }
     },
     "b9c99fae866d44889a3672917fec0e21": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "DescriptionStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "DescriptionStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "description_width": ""
      }
     },
     "bffa8c42220a42f6804642d9265b0a2a": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "FloatProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "FloatProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_eef4990c9b6047cda82cf76da55e8ed3",
       "max": 346319111.0,
       "min": 0.0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_acb8645a19cf419d9d4d2cfdac027622",
       "value": 346319111.0
      }
     },
     "eef4990c9b6047cda82cf76da55e8ed3": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     }
    },
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}